The Structural Dynamics of
Soluble and Membrane Proteins
Explored through Molecular Simulations
Dissertation
zur Erlangung des akademischen Grades
Doktor rerum naturalium (Dr. rer. nat.)
von
Samira Yazdi
geboren am 27. Mai 1981 in Mashhad, Iran
genehmigt durch die Fakultät für Verfahrens- und Systemtechnik
der Otto-von-Guericke-Universität Magdeburg
Promotionskomission:
Dr. rer. nat. Matthias Stein
Prof. Dr. rer. nat. Michael Naumann
Prof. Dr. rer. nat. habil. Helmut Weiß
eingereicht am: 3 Mai 2016
Promotionskolloquium am: 27 September 2016
To mom and dad
Contents
Abstract ......................................................................................... i
Kurzfassung ..................................................................................... ii
Abbreviations and Acronyms ................................................................ iii
A. Introduction ................................................................................. 2
B. Protein Structure ........................................................................... 6
B.1 Primary structure .................................................................................................................... 6
B.2 Secondary structure................................................................................................................ 7
B.2.1 The α, 310, and π helices .................................................................................................... 8
B.2.2 The β sheet and β turn ..................................................................................................... 9
B.2.3 Loops ............................................................................................................................... 9
B.3 Tertiary and quaternary structures ........................................................................................ 10
C. Repeat Proteins........................................................................... 12
C.1 Ankyrin repeats ..................................................................................................................... 12
C.2 NF-Bs and their ankyrin repeat inhibitors ........................................................................... 13
C.2.1 NF-B proteins .............................................................................................................. 13
C.2.2 The IB family................................................................................................................ 14
C.3 The canonical NF-B activation pathway .............................................................................. 14
C.4 Medical relevance of malfunctioning in the NF-B/IB complex......................................... 15
C.5 Phosphorylation meets the degron recognition motif........................................................... 15
C.6 IB structural map .............................................................................................................. 16
C.7 Free IB .............................................................................................................................. 16
C.8 The SRD in IB ................................................................................................................... 16
D. Voltage-gated Ion Channels ............................................................. 19
D.1 The biological membrane ..................................................................................................... 19
D.2 The membrane potential ...................................................................................................... 20
D.3 Voltage-gated potassium channels....................................................................................... 21
D.4 The voltage sensor and voltage sensing mechanism ............................................................ 21
D.5 Modulation of voltage sensor in the Shaker .......................................................................... 22
D.6 Disease and dysfunctional KV channels ................................................................................. 23
D.7 PUFA mediated modulation ................................................................................................. 23
E. Methodology ............................................................................... 26
E.1 Molecular modeling .............................................................................................................. 26
E.1.1 Comparative modeling and threading ............................................................................ 27
E.2 Molecular dynamics simulations ........................................................................................... 28
E.2.1 Theory ........................................................................................................................... 28
E.2.2 Force fields .................................................................................................................... 30
E.3 Free energy calculations........................................................................................................ 31
E.3.1 Sampling along a reaction coordinate ............................................................................ 32
E.3.2 Accelerated weight histogram ....................................................................................... 32
F. Study I ...................................................................................... 35
F.1 Structural modeling of the N-terminal signal–receiving domain of IB ............................... 35
F.2 Methods ................................................................................................................................ 36
F.2.1 Structural modeling ....................................................................................................... 36
F.2.2 System assembly and protocol for MD simulations .........................................................37
F.3 Results and Discussion ...........................................................................................................37
F.3.1 Structural elements in the signal-receiving domain (SRD) ...............................................37
F.3.2 Structural refinement by molecular dynamics simulations ............................................. 40
F.3.3 Conformational change induced in IκBα in its bound form to NF-κB............................... 45
F.3.4 Free IκBα vs bound IκBα ................................................................................................. 48
G. Study II..................................................................................... 53
G.1 Double phosphorylation-induced structural changes in the signal-receiving domain of IB in
complex with NF-κB ................................................................................................................... 53
G.2 Methods ............................................................................................................................... 55
G.2.1 Structural model of IB/NF-B .................................................................................... 55
G.2.2 Molecular simulation setup ........................................................................................... 55
G.3 Results .................................................................................................................................. 56
G.3.1 Local divergence in structural stability promoted by double-phosphorylation ............... 56
G.3.2 Double-phosphorylation induces an extended N-terminal
conformation of SRD .............................................................................................................. 58
G.3.3 Variation in solvent exposure in Ser32 and Ser36 ........................................................... 59
G.3.4 Double-phosphorylation stabilizes region by novel hydrogen
bond interactions.................................................................................................................... 61
G.3.5 Electrostatic effects ....................................................................................................... 62
G.4 Discussion ............................................................................................................................ 63
H. Study III .................................................................................... 67
H.1 The Molecular Basis of Polyunsaturated Fatty Acid Interactions with the Shaker VoltageGated Potassium Channel........................................................................................................... 67
H.2 Methods ............................................................................................................................... 70
H.2.1 Shaker channel system setup in open state.................................................................... 70
H.2.2. PUFA and SFA system setup in open state simulations ................................................ 70
H.2.3 Modeling of Shaker channel in closed state ....................................................................71
H.2.4 In silico mutagenesis of the VSD.................................................................................... 72
H.2.5 Small-scale PUFA and SFA systems ............................................................................... 72
H.3 Results .................................................................................................................................. 72
H.3.1 Saturation levels in the carbon tail affect the structural dynamics ................................. 72
H.3.2 PUFA interactions with the Kv Shaker channel in the open state .................................... 74
H.3.3 PUFA interactions with the Kv Shaker channel in the closed state ................................... 77
H.3.4 SFA interactions with the Kv Shaker channel in the open and closed state ..................... 78
H.3.5 Electrostatic interactions characterize the PUFA interaction sites ................................. 80
H.4 Discussion ............................................................................................................................ 82
I. Study IV ..................................................................................... 85
I.1 The Regulatory Role of Polyunsaturated Fatty Acids on the Free-Energy Landscape of KV
Shaker Channel Deactivation ...................................................................................................... 85
I.2 Methods ................................................................................................................................. 87
I.2.1 Shaker channel in PUFA and PUFA-free systems ............................................................. 87
I.2.2 Accelerated weight histogram calculations ..................................................................... 88
I.3 Results ................................................................................................................................... 88
I.3.1 Sampling approach ......................................................................................................... 88
I.3.2 PUFA affects the free-energy landscape of deactivation ................................................. 89
I.3.3 PUFA stabilizes the open state configuration .................................................................. 90
I.3.4 310-helix integrity ............................................................................................................. 91
I.4 Discussion .............................................................................................................................. 93
Concluding Remarks and Future Perspectives ........................................... 95
Bibliography .................................................................................. 96
Acknowledgements ......................................................................... 115
List of Publications ......................................................................... 116
i
Abstract
From a deeper knowledge of the dynamical nature of biological macromolecules we get an improved
understanding of their life sustaining roles. Computational methods are of absolute necessity in bringing this
about when experimental procedues face their limitations. Throughout this thesis, several computational
techniques have been applied to study the complex nature of protein complexes either in a solvent or a
membrane environment. The two classes of molecules presented in this thesis are the water-soluble proteins
represented by repeat proteins (studies I & II) and membrane proteins illustrated by voltage-gated ion
channels (studies III & IV). The insight gained into the structural characteristics and patterns of interactions of
these proteins is invaluable in drug design.
STUDIES I & II. IκBα, the transcription factor NF-κB inhibitor, is an ankyrin repeat protein that retains NF-κB in
the cytoplasm. Recent protein crystal structures of IκBα in complex with NF-κB have not revealed structural
details about the N-terminal signal receiving domain (SRD), which harbors the sites of post-translational
modifications at sites Ser32 and Ser36. By combining secondary structure annotation and domain threading
followed by molecular dynamics simulation, I have showed that the SRD possesses well-defined secondary
structure elements: 3 additional stable α-helices supplementing the six ankyrin repeats (ARs) present in
crystallized IκBα. Moreover, differences in structural topology and dynamics were observed by comparing the
structures of free and bound IκBα. I also investigated the effect of post-translational mono- and doublephosphorylation of the serine residues of the SRD. Mono-phosphorylation at either Ser32 or Ser36 was not
sufficient to induce significant structural changes in the secondary structure of the SRD of IκBα. Doublephosphorylation yielded a reduced distance between the Cα atoms of these serine residues, inducing an
extended conformation, which renders it accessible by the E3 ligase.
STUDIES III & IV. Polyunsaturated fatty acids (PUFA) have been reported to influence the gating mechanism
by electrostatic interactions to the voltage-sensor domain (VSD) of voltage-gated potassium (KV) channels,
however the exact PUFA-protein interactions have remained elusive. I have reported on the interactions
between the Shaker KV channel in open and closed states and a PUFA-enriched lipid bilayer using molecular
dynamics simulations. A putative PUFA binding site was determined in the open state of the channel located
at the protein-lipid interface in the vicinity of the extracellular halves of the S3 and S4 helices of the VSD. In
particular, the lipophilic PUFA tail covered a wide range of non-specific hydrophobic interactions in the
hydrophobic central core of the protein-lipid interface, while the carboxylic head group displayed more
specific interactions to polar/charged residues at the extracellular regions of the S3 and S4 helices,
encompassing the S3-S4 linker. Moreover, by studying the interactions between saturated fatty acids (SFA)
and the Shaker KV channel, the study confirmed an increased conformational flexibility in the polyunsaturated
carbon tails compared to saturated carbon chains, which may explain the specificity of PUFA action on
channel proteins. I also computed the energetics of the free-energy landscape of the Shaker KV channel in
lipid bilayers free from as well as enriched with PUFAs. By choosing a reaction coordinate along the vertical
translation of S4 towards its down state in the deactivation pathway, I investigated the free energy
differences in passing the first energy barrier of deactivation, i.e. going from the open (O) state to the closed1 (C1) state, in a KV channel that appears to be affected by PUFAs.
ii
Kurzfassung
Durch das tiefere Verständnis der dynamischen Natur biologischer Makromoleküle gewinnen wir eine
verbesserte Einsicht in ihre lebenswichtigen Rollen und Funktionen. Computergestützte Ansätze sind hierbei
eine absolute Notwendigkeit, wenn experimentelle Methoden angesichts ihrer zeitlichen und räumlichen
Begrenzungen nicht ausreichen. In dieser vorgelegten Arbeit wurden verschiedene computergestützte
Techniken angewendet, um die komplexe Art von verschiedenen Proteinkomplexen in Lösung und in einer
Membranumgebung zu untersuchen. Die zwei Klassen von Molekülen in dieser Arbeit sind wasserlösliche
Proteine am Beispiel der Repeatproteine (Studien I und II) und membrangebundene Proteine am Beispiel des
spannungsgesteuerten Ionenkanals (Studien III und IV). Die gewonnenen Einsichten in die strukturellen
Besonderheiten und Wechselwirkungsmuster dieser Proteine ist wertvoll für mögliche
Medikamentenentwicklungen.
STUDIEN I und II. IκBα, der Inhibitor des Transkriptionsfaktors NF-κB, ist ein Ankyrin-Repeat-Protein, welches
NF-κB im Zytoplasma hält. Die neuesten Proteinkristallstrukturen von IκBα im Komplex mit NF-κB haben
nicht die strukturellen Details der N-terminalen signalempfangenden Domäne (SRD) aufklären können, die
die Orte der posttranslationalen Modifikationen an den Positionen Ser32 und Ser36 enthält. Durch eine
Kombination von Sekundärstrukturannotationen und Proteindomänenthreading, gefolgt von
Molekulardynamiksimulationen, habe ich gezeigt, dass die SRD wohl definierte ekundärstrukturelemente
enthält: 3 weitere stabile α-Helices, zusätzlich zu den sechs Ankyrin Repeats (AR) aus dem kristallisierten
IκBα. Des Weiteren wurden strukturelle Topologien und die Dynamiken von Strukturen des freien und
gebundenen IκBα untersucht. Ich habe auch den Effekt der posttranslationalen einfachen und zweifachen
Phosphorylierung der Serinaminosäuren in der SRD untersucht. Die einfache Phosphorylierung entweder am
Ser32 oder Ser36 war nicht ausreichend, um signifikante strukturelle Veränderungen in der Sekundärstruktur
der SRD von IκBα zu induzieren. Die zweifache Phosphorylierung führte zu einem verringerten Abstand
zwischen den Ca-Atomen dieser Serinaminosäureste, induzierte eine gestreckte Konformation des
Proteinrestes und machte diesen so der E3 Ligase zugänglich.
STUDIEN III UND IV. Mehrfachungesättigte Fettsäuren (‚polyunsaturated fatty acids‘, PUFA) wurden in der Literatur beschrieben, die Schaltvorrichtung des spannungsgesteuerten Kalium (K V) Kanals durch
elektrostatische Wechselwirkungen mit der Spannungssensordomäne (VSD) zu aktivieren. Die genaue PUFAProtein Wechselwirkungen konnten bisher nicht aufgeklärt werden. Ich habe die Wechselwirkungen zwischen
dem Shaker KV-Kanal im offenen und geschlossen Zustand und einer PUFA-angereicherten
Lipiddoppelschicht mit Molekulardynamiksimulationen untersucht. Eine mögliche PUFA-Bindungsstelle
wurde im offenen Zustand des Kanals lokalisiert an der Protein-Lipid-Grenzfläche in der Nähe der
extrazellulären Hälften der S3- und S4-Helices der VSD. Der lipophile PUFA-Schwanz zeigte eine große
Anzahl von nicht-spezifischen hydrophoben Interaktionen im zentralen hydrophoben Kern der Protein-LipidGrenzfläche. Die Carboxylat-Kopfgruppe dagegen wies spezifischere Wechselwirkungen mit
polaren/geladenen Resten an den extrazellulären Regionen der S3- und S4-Helices auf, den S3-S4 Linker
umfassend. Durch eine Untersuchung der Wechselwirkungen zwischen gesättigten Fettsäuren (SFA) und dem
Shaker KV-Kanal konnte darüber hinaus eine zunehmende konformationelle Flexibilität der ungesättigten im
Vergleich zu den gesättigten Kohlenstoffketten bestätigt werden. Dies kann die Spezifität der Wirkung von
ungesättigten Fettsäuren (PUFA) auf Kanalproteine erklären. Ich habe ebenfalls die relativen Energien der
freien Energielandschaft des Shaker KV-Kanals in Lipiddoppelschichten in Abwesenheit und angereichert mit
PUFA-Molekülen berechnet. Durch die Wahl einer Reaktionskoordinate entlang der vertikalen Translation
von S4 vom offenen zum geschlossenen Zustand entlang des Deaktivierungspfades konnten die Unterschiede
in freien Energien bei der Überquerung der ersten Barriere der Deaktivierung, d.h. vom offenen (O) zum
geschlossenen-1 (closed-1‘, C1) Zustand in einem KV-Kanal berechnet werden, der betroffen durch PUFAs ist.
iii
Abbreviations and Acronyms
AR
ARD
AWH
C-terminal
DHA
KV
MD
NLS
N-terminal
NMR
PD
PDB
PEST
PIP2
POPC
PUFA
RMSD
RMSF
SCF
SFA
SRD
SSE
TM
VSD
ankyrin repeat
ankyrin repeat domain
accelerated weight histogram
carboxyl-terminal
docosahexaenoic acid
voltage-gated potassium channel
molecular dynamics
nuclear localization sequence
amino-terminal
nuclear magnetic resonance
pore domain
protein date bank
proline glutamate serine threonine
phosphatidylinositol-4,5-biphosphate
1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine
polyunsaturated fatty acid
root mean square displacement
root mean square fluctuation
Skp1-Cullin-F-box
saturated fatty acid
signal receiving domain
secondary structure element
transmembrane
voltage-sensor domain
One and three letter codes for amino acids:
Ala (A)
Arg (R)
Asn (N)
Asp (D)
Cys (C)
Gln (Q)
Glu (E)
Gly (G)
His (H)
Ile (I)
Alanine
Arginine
Asparagine
Aspartic acid
Cysteine
Glutamine
Glutamic acid
Glycine
Histidine
Isoleucine
Leu (L)
Lys (K)
Met (M)
Phe (F)
Pro (P)
Ser (S)
Thr (T)
Trp (W)
Tyr (Y)
Val (V)
Leucine
Lysine
Methionine
Phenylalanine
Proline
Serine
Threonine
Tryptophan
Tyrosine
Valine
Chapter A. Introduction
1
Chapter A
Chapter A. Introduction
2
Introduction
Living organisms are composed of cells, basic structural and functional units of life that
carry out vital functions such as maintenance, recycling, disposing of waste, adapting to its
environment and replication. Cells can exist as uni-cellular organisms such as many species
of bacteria and protozoa, or they can be part of multi-cellular complex units.
Cells are shaped by a structure known as the cell membrane, also referred to as the
plasma membrane, separating the contents of the cell from the surroundings. The plasma
membrane is constituted of phospholipids, fat-based molecules, which prevent hydrophilic
substances from entering or leaving the cell. Embedded within the membrane are proteins
that serve a range of functions. The gatekeepers decide on which substances that are
allowed to diffuse across the membrane; markers identify a cell as friendly and part of the
organism or as a foreign molecule; fasteners bind cells together and enable them to
function as one unit; and the communicators have the purpose of sending and receiving
signals from other cells and the environment.
Within the interior walls of the cells an aqueous solution called the cytoplasm
retains all the cellular machinery and structural elements of the cell. The largest organic
intra-cellular components in the cytoplasm include carbohydrates, lipids, nucleic acids, and
proteins. Carbohydrates expressed in simple and complex forms, are the starches and
sugars in the cell. The simple carbohydrates supply the cell’s immediate energy demands and the complex carbohydrates act as intracellular energy storage. Lipids are the
components of the plasma membrane in addition to many intracellular membranes. They
partake in conveying signals both within the cell and from the outside to the inner cell. A
cell’s genetic code is expressed through nucleic acids of which there are two major classes:
deoxyribonucleic acid (DNA) and ribonucleic acid (RNA). DNA contains the information,
which is necessary to build and maintain the cell, while RNA is involved with the expression
Chapter A. Introduction
3
of the information stored in DNA. An organism’s genome is the complete set of genes that
is made up of DNA and contained within the nucleus. In bacteria and archaea, the nucleus is
not separated from the cytoplasm, whereas in eukaryotes, the nucleus contains nuclear
material enclosed within a double membrane, the nuclear envelope. The information
programmed into the nucleotide sequence of a cell’s nucleic acids determines the structure of every protein and essentially of every molecule and cellular component. The flow of
genetic information in cells from DNA encoding RNA and RNA encoding protein is known
as the central dogma of life (Figure A.1) [1].
Figure A.1 The central dogma of life: from DNA, to RNA, to protein. Figure adapted from
[2].
Proteins serve a variety of functions in the cell, both catalytic and structural and are
made from chains of smaller molecules called amino acids. The malfunctioning of proteins
leads to a great many illnesses, ergo understanding their structure and thus function is
central to deciphering biology and life. In 1958, John Kendrew and colleagues showed for
the first time the three-dimensional structure of a protein, the low-resolution structure of
myoglobin, a monomeric hemeprotein [3]. This was followed two years later, in 1960, by
the 5.5 Å resolution structure of haemoglobin, a tetrameric hemeprotein, solved by Max
Perutz and co-workers which was the first indication of protein families [4]. This marked
only the beginning of a fascinating newborn world of protein science where atomic-level
resolution structures came to give way for a new understanding of protein structure and
their intricate function.
The aim of this thesis has been to study proteins, their structure and interactions,
and concludes the results and findings of my projects during my PhD studies. I have studied
both soluble and membrane proteins employing computational techniques such as
molecular modeling, molecular dynamics simulations, and free-energy calculations. In this
thesis I will present the obtained results on four different studies. In study I, to better
understand the structural aspects of the NF-κB activation pathway, I have modeled the
three-dimensional structure of the N-terminal region of the NF-κB inhibitor, IκBα. Study II
has been the focus of posttranslational modification, in particular phosphorylation, and its
structural effects on IκBα. In study III, I change focus over to membrane proteins and
investigate the molecular details of the interactions between the Shaker voltage-gated
Chapter A. Introduction
4
potassium channel and polyunsaturated fatty acids. Finally, study IV deals with mapping
the free-energy landscape of polyunsaturated fatty acid modulation on the Shaker voltagegated potassium channel.
Chapter B. Protein Structure
5
Chapter B
Chapter B. Protein Structure
6
Protein Structure
B.1 Primary structure
The primary structure of proteins is the sequence of amino acids, which forms a chain and
determines the fold of a protein. Each amino acid is specified by a nucleotide sequence.
Nucleotides are the building blocks of DNA that contain information for the synthesis of
RNA or protein. There are 4 major bases in DNA: adenine (A), guanine (G), cytosine (C), and
thymine (T) and these are linked covalently together via a phosphodiester linkage. A triplet
of nucleotides, named a codon, codes for a specific amino acid. The three letter code gives
43 = 64 combinations of amino acids with most of the 20 natural amino acids having more
than one codon.
Every amino acid has a hydrogen, carboxyl group, and an amino group bound to a
central Cα in addition to a side chain group. This side chain group varies in structure, size,
and electric charge between the different amino acids. Classification of amino acids based
on the properties of the side chain results in five major classes: nonpolar aliphatic,
aromatic, polar uncharged, positively charged, and negatively charged. Amino acids come
together to form proteins by joining covalently through an amide linkage, named a peptide
bond. The formation of the peptide bond occurs through a condensation reaction in which
the carboxyl group of one amino acid reacts with the amino group of another resulting in
the production of a water molecule.
Several amino acids join together to form a peptide, and once incorporated into a
peptide chain, amino acids are referred to as residues. A peptide chain has two free ends:
an amino-terminal (N-terminal) end which is the amino acid residue at the end with a free
amino group, and the carboxyl-terminal (C-terminal) end which is the residue at the other
Chapter B. Protein Structure
7
end with a free carboxyl group. Protein chains built up of less than 50 amino acids are
referred to as peptides, and when longer either polypeptide or ‘protein’ [1].
B.2 Secondary structure
The secondary structure of a protein describes stable arrangements of amino acid residues
resulting in recurring structural patterns. Three dihedral angles, ϕ (phi), ψ (psi), and ω
(omega), known as torsion angles, define protein formation by rotation about each bond in
the protein backbone.
ϕ involves the C-N-Cα-C bonds, with rotation occurring about the N-Cα bond; ψ
involves the N-Cα-C-N bonds, with rotation occurring about the Cα-C bond; and ω involves
the Cα-C-N-Cα bonds, with rotation occurring about the C-N bond. The ϕ and ψ angles are
defined as ±180°, but the peptide bond ω is not free to rotate. In theory, the ϕ and ψ angles
can take any value in the –180° and +180° range, but because of steric interference
between the atoms in the protein backbone and amino acid side chains there is a restriction
on the allowed angles.
The Ramachandran plot is a two-dimensional map of the allowed ϕ and ψ angles in a
protein, based on calculations using known van der Waals radii and dihedral angles (Figure
B.1) [5,6]. When the torsion angles ϕ and ψ remain the same throughout a segment they
shape a regular secondary structure. Two of the main secondary structure folds proposed
by Pauling et al. are the α helix and β sheet [7-9]. Other commonly found secondary
structure elements include the 310-helix, π-helix, β-turn, loops and coils. In addition to being
the foundation of tertiary structure, secondary structure folds play other roles such as
allowing efficient packing of atoms, contributing to protein stability, and forming
biologically active structures.
Chapter B. Protein Structure
8
Figure B.1 Ramachandran plot. The shaded areas in light and dark green represent the
allowed and favored regions for backbone dihedral angles ψ against ϕ. The pink regions
correspond to regions allowing α-helical (left) and β-sheet conformations (right). Figure
adapted from [10].
B.2.1 The α, 310, and π helices
The α-helix structure is the simplest arrangement a polypeptide chain can adopt given its
rigid peptide bonds. The helical structure can be between 5 to 40 amino acid residues,
corresponding to backbone ϕ and ψ angles of -57° and -47°, respectively and each helical
turn including 3.6 amino acid residues [1]. It happens that these structures deviate from the
dihedral angles giving rise to bends or kinks in the helical structure.
The α-helix is a right-handed helix that is stabilized by an extensive network of
hydrogen bonds between the peptide bonds, where each successive turn of the α helix is
held to adjacent turns by three or four hydrogen bonds (Figure B.2). In specific, each bond
is formed between a backbone carbonyl group (C=O) and a backbone amide group (N-H) of
the fourth amino acid located down the amino-terminal side of the peptide. Hydrogen
bonds are not the only main stabilizing factor in the helix, nonpolar and van der Waals
interactions between the residue side chains in particular the Cα atom also contribute to the
stability of the helix [11,12]. Since the helical conformation is the most compact out of all
secondary elements, amino acid residues that have smaller and more linear side chains are
more preferred since there is less chance for steric interference. The most commonly
occurring amino acids in α-helices are Ala, uncharged Glu, Leu, Arg, charged Met, and
charged Lys.
Chapter B. Protein Structure
9
Figure B.2 Helical conformations. Comparison of the α-helix (left) and the 310-helix (right).
The 310-helix is tighter wound with residue i interacting with residue i+3, versus the i+4
interaction in the α-helix. Figure adapted from [13].
The 310-helix is narrower and longer than the α-helix, including 3 residues and 10
backbone atoms per turn, hence the origin to its name. The 310-helix is a rarely observed
secondary structure element in particular because of the i  i + 3 hydrogen bond pattern
which renders this helix energetically unfavorable due to the poor backbone dipoles
arrangement.
The π-helix on the other hand is shorter and wider than the α-helix, with 4.4
residues per turn. The π-helix is also an energetically unfavorable structural element
because of its i  i + 5 hydrogen bond pattern. Both the 310 and π-helix mainly appear at
the edges of α-helices, consisting of no more than a few residues.
B.2.2 The β sheet and β turn
Contrary to the compact structure of the α-helix, the β-strand adopts a less compact
structure with the amino acid residues preferring conformations with ϕ = -120° and ψ =
120°. β-strands typically have between 5 and 10 residues, with the distance between the Cα
atoms of the adjacent residues over twice as long as in the α-helix, thereby making the βstrand a more extended conformation. The extended conformation of the β-strand does
not permit intermolecular hydrogen bonds within the same strand, or stabilizing van der
Waals contacts. Consequently, its is rare that β-strands appear in isolation and instead are
found to form flat structures arranged alongside one another, known as the β-sheet.
There are two types of β-sheets, the anti-parallel and the parallel. In the anti-parallel
β-sheet the strands are separated from one another by a few residues, that form a loop
known as the β-turn, allowing the β-strands to line side by side in opposite directions.
Conversely, in the parallel β-sheet, the residues connecting each β-strand are long enough
to allow for the arrangement of the strands in the same direction in a parallel fashion.
The hydrogen bond network in β-sheets is between strands where most of the
backbone groups are involved in hydrogen bonds. The preferred amino acids in β-sheets
are Asp, Asn, Ser, and Pro which appear in their C-termini [14], whereas the amino acid
preference in the N-termini are Lys and Arg. In general, β-sheets are more accommodating
to residues with bulkier side chains, owing to their less compact structure.
B.2.3 Loops
The segments of proteins that lack a regular secondary structure are referred to as loops.
This absence of secondary structure makes loops more flexible, which also means that the
loop backbones are not involved in a hydrogen bond network like helical and β
conformations [15]. Loops consist mainly of polar residues, and are more prone to appear
on the surface of the protein as opposed to the hydrophobic core.
Chapter B. Protein Structure
10
B.3 Tertiary and quaternary structures
Tertiary structure refers to the three-dimensional folding of proteins, which normally
determines the function of a protein. Most of the interactions between the amino acids in a
protein are non-covalent, i.e. van der Waals, electrostatic or nonpolar [16]. This is
particularly important since the large number of interactions contributes towards
stabilizing the tertiary structure of proteins, but since these interactions are relevantly
weaker than covalent bonds, protein structures remain dynamic rendering them
indispensable to their functions. In the core of the protein that is mainly built of nonpolar
residues, it is primarily nonpolar interactions that are the strongest. However, on the
surface of the protein, it is mostly electrostatic interactions that are prevalent, where polar
residues interact with each other or the surrounding water molecules and ions.
The structural arrangement of proteins with more than one independent subunit is
described as quaternary structure. When the subunits are identical the protein is a
homomer, and when they are different it is called a heteromer. Quaternary structures are
stabilized mainly by non-covalent interactions, and the formation of these structures is
driven by the hydrophobic effect [17].
Chapter C. Repeat Proteins
11
Chapter C
Chapter C. Repeat Proteins
12
Repeat Proteins
Repeat proteins are non-globular folds that are built from repeated motifs of between 20
and 40 amino acids stacked together to form extended super-helical structures [18,19].
These proteins differ from globular proteins by their repetitive and elongated structures
that are governed mainly by local and short-range interactions. Examples of a few repeat
protein families with different motifs are: β strand – link – β strand motif such as
hexapeptide (HPR) and WD40 repeats, α-helix – link – α-helix motif such as TPR and HEAT
repeats, α-helix – link – β-strand motif such as leucine-rich (LRR) repeats, α-helix – β-hairpin
or loop – α-helix motif such as ankyrin (ANK) repeats [20].
No single motif is yet known to form a stable folded unit, as stacks of these
structural units are needed to form stable folded structures. What unifies different repeat
proteins is their ability to mediate protein-protein interactions. By forming scaffolds, there
proteins are able to bind varying binding partners, thereby making these proteins open to
varied cellular functions such as protein transport, protein complex assembly, and protein
regulation. Next, I will turn my focus to ankyrin repeats and discuss in further detail the
nature of one of the most frequently observed motifs.
C.1 Ankyrin repeats
The ankyrin repeat (AR) is present in about 6 % of eukaryotic protein sequences, making it
one of the most common sequence motifs. The ankyrin repeat is a 33 residue sequence
motif that was first discovered in the yeast cell cycle regulator Swi6/Cdc10 and the
Drosophila signaling protein Notch [21] and takes its name from the human erythrocyte
protein ankyrin [22]. While ankyrin repeats are found in all three kingdoms including
Chapter C. Repeat Proteins
13
bacteria, archaea, and eukarya in addition to a few viral genomes, the majority of these
repeats are present in eukaryotes. These modular protein domains, which act as scaffolds
for molecular interactions, are deemed crucial for the development of many signaling
pathways [23].
The PFAM and SMART databases reveal proteins containing up to 33 ankyrin
repeats with the majority having six or fewer repeats. Identification of terminal repeats has
proven to be a challenge because of the divergence of many well-conserved hydrophobic
residues. Terminal repeats often deviate from the customary consensus sequence because
these positions have through evolution been replaced with polar residues to enhance
favorable interactions with the solvent. Also, as a result of truncation, terminal repeats are
often not accounted for in protein sequences. This opens up the possibility of detected
ankyrin repeat proteins having one or two more additional repeats in their sequences in
reality. Ankyrin repeats can either exist as single protein domains or in coexistence with
other domains in the same protein [24].
Individual ankyrin repeats fold into two antiparallel α-helices followed by a β-hairpin
or a long loop. The repeats are assembled in an array and form an L-shaped domain,
resembling a cupped hand with the fingers represented by the hairpins and the palm by the
helices [20,25]. Structures of proteins with many ankyrin repeats reveal an overall curved
shape. Ankyrin repeats are quite well-conserved and rarely deviate from the canonical
motif, with the exception of insertions mainly in the loop areas. The helices in the repeats
are primarily stabilized by hydrophobic interactions, whereas the β-hairpins and the loops
are connected by a hydrogen bonding network. The inner helix is seven residues long
extending from positions 5-12 in the canonical sequence, while the outer helix being nine
residues long spans the positions 15-24. In positions 4-7 is the prevailing TPLH motif, with
the proline causing a tight turn leading to the L-shaped structure of the ankyrin repeat
[24,25].
The ankyrin repeat is present in many biologically important proteins that are
implicated in protein functions such as cell-cell signaling, cytoskeleton integrity,
transcription and cell-cycle regulation, inflammatory response, and development. A few
families containing ankyrin repeats include the INK4 tumor suppressor family, Notch the
signaling protein involved in development, the TRP cation channel family, and NF-B, the
transcription factor regulating inflammatory response inhibited by the ankyrin repeat
protein IBα. Before we shift our attention to the ankyrin repeat protein IBα, we will first
take a look at the NF-B activation pathway and understand its significance.
C.2 NF-Bs and their ankyrin repeat inhibitors
C.2.1 NF-B proteins
Initially NF-B was identified as a transcription factor binding to the enhancer element of
the immunoglobulin (Ig) light-chain gene in B cells [26,27]. However, it later became
apparent that it is a major regulator of innate and adaptive immunity and inflammatory
responses present in every cell type, retained in the cytoplasm in an inactive mode bound
to the IB inhibitors [28,29]. When in an active DNA-binding mode, NF-B is a heterodimer
Chapter C. Repeat Proteins
14
composed of a combination of members from the NF-B/Rel family. The known members
of this family are: NF-B1 (p50 and its precursor p105), NF-B2 (p52 and its precursor p100),
c-Rel, RelA (p65), and RelB [30]. A well-conserved 300 amino acid Rel homology region
(RHR), composed of two immunoglobulin domains, is common between these proteins.
The RHR region is involved in DNA binding, dimerization, and interaction with the IB
proteins, and contains a nuclear localization sequence (NLS) [31]. Knockout studies in mice
have identified the p65/RelA protein to be crucial for survival. The p65/p50 heterodimer is
the most occurring NF-B member throughout different cells and is the protein complex in
question when referring to NF-B [30,32].
C.2.2 The IB family
The members of the IB family include IB, IB, IB, IB, Bcel-3, p105, p100, and the
Drosophila protein Cactus [30,33]. The IB members are ankyrin repeat proteins each
containing six or seven ankyrin repeats which enable binding to the RHR and masking of
the NLS in NF-B. The most vital NF-B regulators are IB, IB, and IB all of which
have a regulatory N-terminal region that is required for stimulus induced degradation, an
important step in NF-B activation [34].
One of the first and best-characterized IB family members is IB [35]. Upon
synthesis, IB enters the nucleus and binds to NF-B enabling its dissociation from DNA,
leading it back to the cytoplasm through its nuclear export sequence (NES) [36]. The three
classical IBs experience signal induced proteasomal degradation with varying kinetics [37].
As opposed to much slower kinetics in IB and IB degradation, IB displays higher
kinetics in response to inflammatory stimuli such as TNF- and lipopolysaccharide (LPS)
and via a negative feedback loop, synthesized IB enters the nucleus to bind to
deacetylated p65/p50 dimers and carry them back to the cytoplasm [36,38].
C.3 The canonical NF-B activation pathway
The most well known NF-B activation pathway is the so called ‘canonical’ pathway, which
through a variety of stimuli activates NF-B and in turn induces the rapid degradation of
IB and causing the nuclear accumulation of NF-B. One of the primary and critical events
in this activation pathway is the phosphorylation of IB at serines 32 and 36 in the Nterminal regulatory domain by the IB kinase IKK [39,40].
IKK consists of two catalytic kinases, the helix – loop - helix (HLH) serine-threonine
kinases, IKK and IKK [40,41], and a noncatalytic signal recognition scaffolding protein
known as IKK [42], the IKK associated protein IKAP [43], and FIP the 14.7 interacting
protein with the role of a signaling adapter and scaffolding protein. IKK activation is
identified as the main regulatory step in the canonical pathway since IKK is known to have
a 30-fold higher activity toward IB [41,44] and its genetic deficiency blocks IB
proteolysis by various inducers [45]. IKK is triggered to interact with its substrate through
its activation by phosphorylation of two activation loop serines [46].
Phosphorylation of IB leads to the substrate being recognized by the Fbox/WD40 E3RSIB/-TrCP that initiates the polyubiquitination of IB at lysines 21 and 22
Chapter C. Repeat Proteins
15
by the Skp1-Cullin-F-box(SCF)-type E3 ubiquitin ligase which results in the rapid
degradation of IB through the 26S proteasome [35,47-49]. Upon degradation of its
inhibitor, the free NF-B NLS is exposed and targeted for binding to karyopherins and its
translocation to the nucleus [50].
C.4 Medical relevance of malfunctioning in the NFB/IB complex
Misregulation of the NF-B activation pathway has implications in many chronic diseases
such neurodegenerative diseases, auto-immune diseases and cancer [51-53], making it a
suitable therapeutic target. Any of the events of the signaling pathway can be a target for
therapeutic intervention whether it is the phosphorylation, ubiquitination, or the
degradation of the IB proteins.
C.5 Phosphorylation meets the degron recognition
motif
In resting cells, the half-life of a stable IB is 138 min, but upon phosphorylation of resides
32 and 36, the half-life of IB in these stimulated cells is reduced to 1.5 min [54].
Phosphorylation of IB induces recognition by the SCF-type E3 ligase. The E3 recognizes a
variety of different substrates allowing for specificity but only phosphorylated substrates
through its F-box component, the variable receptor subunit [55]. The SCF complex
thereafter forms a bridge between the substrate IB and the E2 ubiquitin-conjugating
enzyme, which enables the ubiquitination of IB at lysine sites 21 and 22 [56].
The F-box domain of E3, -TrCP, was identified by mass spectroscopy and interacts
specifically with a phosphorylation-based motif through its WD40 domain [57]. This
phosphorylation-based motif, DpSGXpS, is also known as the degron motif ( represents
a hydrophobic residue and X stands for any residue).
The two best characterized substrates of -TrCP are IB and -catenin both
sharing the degron motif. It was Pavletich and colleagues that shed light on the structural
basis for recognition of a phosphorylated IB by the E3 ligase by solving the crystal
structure of -TrCP in complex with the phosphopeptide -catenin [56]. The interacting
face of -TrCP is the -propeller shaped by the seven WD repeat domain of -TrCP where a
central groove going through the -propeller bonds with the degron-containing segment of
the substrate. The three residues accommodated between the two phosphoserines insert
the farthest into the groove making intermolecular contacts in a rather buried
environment. The two phosphorylated serines on the other hand bind at the rim of the
groove through hydrogen bonds and electrostatic interactions.
Chapter C. Repeat Proteins
16
C.6 IB structural map
IBs contain a signal receiving domain (SRD), six to seven ankryin repeat units [58] and a
largely unstructured PEST (enriched in amino acids proline (P), glutamate (E), serine (S) and
threonine (T)) domain at the C-terminus. The C-terminal PEST domain is also the site of
post-translational modifications due to the casein kinase II (CK2) phosphorylation at
positions 283, 288, 291, 293 and 299 [59,60].
The two crystal structures of IB, determined by Huxford et al. [61] at 2.3 Å and by
Jacobs and Harrison [62] at 2.7 Å resolution, illustrate how the six repeating ankyrin domain
assumes the shape of an arched cylinder assembled on top of the interface of the NF-B
heterodimer. Every repeat unit in IB is composed of two -helices connected to each
following repeat with a loop of varying size and a -hairpin turn containing short -strands.
However, repeats 1, 3, and 4 deviate from the canonical 33 amino acid repeat unit. These
repeats are longer than the repeat units in the ankyrin consensus sequence, with the
insertions contained in the loop sections, as these regions are those with the lowest
sequence similarity among all ankyrin repeat proteins. Lack of homology is also observed in
the sixth and last repeat unit, where the dissimilarity falls in after the second helix clearing
the last 11 residues of any secondary structural elements [61].
C.7 Free IB
Free IB (67-317) was characterized by circular dichroism (CD) spectroscopy, 8-anilino-1napthalenesulphonic acid (ANS) binding, differential scanning calorimetry (DSC), and
amide hydrogen/deuterium exchange experiments [63]. The CD spectrum of free IB is
nearly identical to the CD spectrum of the IB/NF-B complex but it shows significant
ANS binding and rapid amide exchange over much of the protein. These findings suggest
that the secondary structure of IB is formed but the tertiary structure may not be
compact. The -hairpins of AR2 and AR3 were remarkably resistant to exchange, whereas
AR5 and AR6 exchanged completely within the first minute in free IB. When bound to
NF-B, the -hairpins of AR5 and AR6 showed dramatically less exchange in the bound
state [64].
C.8 The SRD in IB
The SRD of IB is the central signal receiving and transmitting domain when activating
NF-κB. It contains sites for post-translational modifications (phosphorylation by kinases
IKKα and IKKβ [61] at Ser32 and Ser36; and Lys21 and Lys22 as the sites for subsequent
ubiquitination by SCF(β-TrCP), respectively [62,65]). The SRD was always assumed to be
unstructured or highly disordered based on the failed attempts to crystallize full-length
IκBα in complex with NF-κB. The instability of free IκBα in solution and the absence of
significant SRD contributions to the interaction energy of the protein-protein complex of
IκBα/NF-κB lead to the hypothesis of the SRD not being critical for this complex formation.
Chapter C. Repeat Proteins
17
Detailed knowledge of the NF-κB/IκBα interaction comes from protein
crystallography and high resolution NMR experiments [66]. However, these results do not
include any structural information about the SRD (residues 1-72) of IκBα. Previous
investigations by molecular dynamics (MD) simulations of NF-κB/IκBα focused on the
amide proton/deuterium exchange kinetics of four central ankyrin repeat units of cocrystallized IκBα by accelerated molecular dynamics (aMD) simulations [65], a truncated
free-IκBα [67] and the structure of a free, doubly phosphorylated 24 amino acid peptide of
the SRD [68].
In studies I and II of this thesis, I have explored the nature of secondary structure
elements in the SRD of IκBα in addition to studying the structural effects that arise as a
response to phosphorylation on this ankyrin repeat protein.
Chapter D. Voltage-gated Ion Channels
18
Chapter D
Chapter D. Voltage-gated Ion Channels
19
Voltage-gated Ion Channels
D.1 The biological membrane
The cell membrane is the permeable structure that envelops a cell and acts as a protection
barrier between the inside, the cytosol, and outside, the extracellular environment,
deciding what enters and leaves the cell. The membrane consists of phospholipids,
proteins, and carbohydrates. The phospholipids form the surfaces of the membrane bilayer
with their polar head groups and their hydrophobic acyl chains tuck in between and shape
the lipid bilayer. The proteins crowding the surface of membranes are peripheral proteins,
whereas the proteins embedded into the membrane bilayer are knows as integral
membrane proteins, which act as channels and transporters for ions and hydrophilic
substances. Membrane bilayers are highly dynamic and fluid structures, which enable
interaction among proteins and between proteins and lipids. The permeability property of
the membrane enables it to maintain charge and concentration gradients crucial to the
function of the cell such as ATP synthesis, flow of solutes across the membrane, and in
nerve and muscle cells produce and transmit electrical signals.
Because of difficulties with purification and crystallization of membrane proteins,
the atomic structure of less than 10 % of known membrane proteins has been determined
so far. The transmembrane (TM) peptide chains of integral membrane proteins are largely
composed of nonpolar residues as these are the segments that are in contact with the
hydrophobic environment of the lipid bilayer, interacting with the hydrophobic acyl groups
of the membrane phospholipids and often take the secondary structure shape of -helices.
A common membrane protein in eukaryotic cells are ion channels; channels that are
characterized by ion specificity and gating and respond to ligands or voltage across the
Chapter D. Voltage-gated Ion Channels
20
membrane bilayer. The characteristics of ion channels have been better understood in part
because of the high-resolution structures of voltage-gated potassium channels [69].
D.2 The membrane potential
A neuron is an electrically excitable cell that processes and transmits information by
electro-chemical signaling and is the principal element of the nervous system. Signals
coming in from other neurons are normally received through the dendrites, and outgoing
signals flow out through the axon of a neuron. The rapid communication between neurons
is achieved through electrical signals. Communication then takes place at synapses through
transmission of chemicals known as neurotransmitters.
It was in 1952 when the first quantitative description of the electrical events,
forming the basis of action potential generation, was proposed by Alan Hodgkin and
Andrew Huxley - using data from the squid giant axon, they proposed a model that
accurately predicted the shape of the action potential [70]. When at rest, neurons generate
a constant voltage across their membranes called the resting membrane potential that is
typically between -40 to -90 mV. In response to stimuli, the resting membrane potential is
changed and electrical signals are produced by the neurons. The electrical signals that are
propagated through their axons are called action potentials [71]. The resting membrane
potential is maintained via an unequal distribution of ions across the membrane that is
created by the Na+/K+ - ATPase.
These ATP-dependent pumps exchange internal Na+ for external K+, consequently
leading to a concentration of K+ ions inside the neuron and Na+ ions outside. A high
concentration of Na+ on the outside of the neuron creates a concentration gradient,
causing Na+ to diffuse into the neuron creating a net positive electrical charge. The influx of
Na+ into the neuron makes the inside more positive - this is known as depolarization and
leads to a reduction and consequently reversal of the transmembrane potential. As the
electrical gradient across the membrane is reduced the influx of Na + ions to the neuron is
diminished. When there is no flow of Na+ ions inside the neuron, an equilibrated state of
membrane potential, known as equilibrium potential, is reached. The equilibrium potential
is deduced from the Nernst equation and for Na+ and K+ lies around +55 mV and -103 mV,
respectively [72].
The action potential can be sectioned into three different stages. At the first stage,
the depolarization stage, the membrane potential rapidly goes from -60 to +40 mV. In the
second stage, the repolarization stage, the membrane potential returns to a resting
potential. In the third stage, known as the after-hyperpolarization, the membrane potential
slowly recovers from the resting potential [72]. The generation of electrical signals in
neurons is conveyed through the concentration gradient maintained by ions. The
membrane proteins that create and maintain ion gradients are known as active
transporters, and the membrane proteins that selectively permeate ions across the
membrane are called ion channels. The ion channels that are able to sense the electrical
voltage across the membrane are called voltage-gated ion channels, which open and close
in response to the membrane potential [71].
Chapter D. Voltage-gated Ion Channels
21
D.3 Voltage-gated potassium channels
Voltage-gated ion channels are the third largest group of signaling molecules encoded by
the human genome, after protein kinases and G protein-coupled receptors [73]. K+ channels
constitute about half of this superfamily which has 78 members. The K+ channel family is
divided into four structural types depending on their mode of activation and the number of
transmembrane segments: the inwardly rectifying 2-transmembrane K+ channels (Kir), 2pore-4-transmembrane K+ channels (K2P), Ca2+activated 6-transmembrane/7transmembrane K+ channels (KCa), and the largest group in this family the voltage-gated 6transmembrane K+ channels (KV) [74]. The KV channel group is encoded by 40 genes in the
human genome and is divided into 12 subfamilies, known as KV1 to KV12 [75]. KV channels
are normally expressed together with either or both voltage-gated Na+ and Ca2+ (CaV)
channels in excitable cells such as neurons or cardiac myocytes, and are responsible for
repolarization after action potential firing, and both voltage-gated Na+ and Ca2+ channels
have been found to be homologous to the KV channels [74].
It was in 1987 that the first structure of a voltage-gated K+ channel, the Drosophila
Shaker channel, was cloned and expressed [76]. Since then, X-ray crystallographic
structures have revealed the atomic structures of several K+ channels, shedding light into
the basic mechanism of K+ channel function. The Streptomyces lividans K+ channel (KcsA)
solved by MacKinnon has contributed greatly into mapping the structural details of a K +
channel [77].
The K+ channel is a tetrameric channel with four identical subunits shaping a central
water-filled ion-conducting pore. Each subunit has six transmembrane -helices, S1-S6,
with both N- and C- termini on the intracellular side of the membrane. The selectivity filter,
which is the formed by the narrowest part of the pore, is shaped by a loop connecting S5
and S6. The pore domain (PD) is formed by S5, a pore helix (PH), and S6 and helices S1-S4
comprise the voltage sensor (VSD). All K+ channels have the signature sequence,
TMxTVGYG, between the two most carboxy-terminal transmembrane helices, the pore
loop [78]. The selectivity filter has a diameter of 3 Å and is highly hydrophilic, made up of
twelve backbone carbonyl groups, three from each of the subunits. KV channels have
another conserved domain, the tetramerization domain (T1) at the N-terminus [79,80],
which determines the specificity of subunit assembly [81] in addition to acting as an
anchoring point of potential –subunits [79].
D.4 The voltage sensor and voltage sensing mechanism
Next to ion permeation as a main function, speed and specificity are two other important
factors when KV channels move ions through their pores. Ion permeation is regulated by
the opening and closing of the pore, which lead to conformational changes known as
gating. This gating is coupled to a sensing mechanism detecting changes in the membrane
voltage. As depolarization prompts the opening of the channel and hyperpolarization
induces its closing, the channel requires a voltage-sensor to detect the potential across the
membrane.
The architecture of an activated (open or slow inactivated) chimaeric KV channel
called the paddle-chimera, in which the voltage sensor paddle has been transferred from
Chapter D. Voltage-gated Ion Channels
22
KV2.1 to KV1.2, and crystallized in complex with lipids at 2.4 Å resolution, has revealed the
pore and voltage sensors embedded in lipids mimicking a membrane bilayer environment
[82]. The VSD is composed of helices S1-S4 that are arranged in an anti-parallel fashion.
The VSD is coupled to the S6 gate through the S4-S5 linker helix and it is this coupling that
enables channel opening and closing. The S4 helix contains the gating charges, five positive
residues mainly arginines thought to be involved in voltage gating of which the first four are
suggested to be the most important – this marks S4 as the voltage sensor [83,84]. The
positively charged residues in S4 interact with negatively counter charges in S1-S3 within
the membrane, while the charged residues exposed to the membrane interact with the lipid
head groups [82,85,86]. This leads to a neutral interior of the channel. A highly conserved
phenylalanine in S2 forms a barrier for the voltage sensor charges [87].
As no crystal structure on the closed state of voltage-gated K+ channels has yet
been solved, the closing of the channel is explained by a few potential models, which
describe the voltage sensor movement. At negative membrane voltages, S4 is trapped in a
down position closer to the intracellular side of the channel. At positive membrane
potentials, S4 is driven in an outward direction through the membrane towards the
extracellular side of the channel. This concerted motion opens the inner gate of the pore
allowing ions to pass through the channel.
Activation can be divided into two main parts: the independent outward movement
of the four S4 helices in a channel, followed by a concerted opening step when all four S4
helices move together [88-92]. However, exactly how S4 moves can be explained by three
main theories, as reviewed by Börjesson and Elinder [93]. The first suggested model is the
helical-screw or sliding helix, which proposes that the positive charges in S4 make contact
with negative counter charges in other parts of the channel as S4 moves 4.5 Å and rotates
60 along its length axis [94,95]. The second model, the transporter-like, helical-twist, or
rocking-banana model, suggest a large rotation but without little translational movement
required to transfer charges from the intracellular to the extracellular side [96]. The third
model is the paddle model, which assumes that the voltage sensor paddle, i.e. S4 and S3b,
are in close contact and never cease this contact during gating [97].
D.5 Modulation of voltage sensor in the Shaker
Ion channel modulation can be achieved by a direct effect on the ion-conduction pore, such
as local anesthetics and the neurotoxins tetrodotoxin from the pufferfish. Another way of
channel modulation is by affecting the channel’s voltage-dependence. By moving the
voltage dependence of a depolarized activated channel in a positive direction along the
voltage axis, the probability of an open channel is decreased, however if you move it in a
negative direction the probability of an open channel is increased [93]. There are a number
of mechanisms that can induce this shift.
Modulation can occur from the intramembrane side and directly affect the voltage
sensor and the gate. The membrane phospholipids for example interact with the channel
by being in direct contact with S4 [98], or the phospholipid phosphatidylinositol-4,5biphosphate (PIP2) that is located in the cytoplasmic leaflet of the membrane and acts
electrostatically on the cytoplasmic parts of the channel mediating its effect [99].
Chapter D. Voltage-gated Ion Channels
23
The voltage dependence of ion channels can also be modified from the extracellular
side through a direct effect on the voltage sensor. For example there are toxins that bind to
the extracellular VSD and trap the voltage sensor either in a resting or an activated state
[100,101]. The tarantula toxin, hanatoxin, has been found to bind to the S3-S4 linker of KV
channels and by preventing the movement of S4 shift the voltage dependence towards
more positive voltages [101,102]. Another type of channel modifier are free fatty acids,
charged lipophilic molecules that interact directly or electrostatically with the channel by
inserting into the membrane near the channel or into hydrophobic cavities on the channel
itself, such as the polyunsaturated fatty acids (PUFAs) [103-108].
D.6 Disease and dysfunctional KV channels
KV channels are important therapeutic drug targets because of their implications in diverse
diseases ranging from cancer to autoimmune diseases to metabolic, neurological and
cardiovascular disorders [74]. In specific, the dysfunction of KV channels has been linked to
epilepsy, a brain disorder that is characterized by recurrent seizures caused by a
synchronous firing of neurons [109]. When antiepileptic drugs do not respond in about 2530 % of children with epilepsy [110], a ketogenic diet has proven to be an effective
treatment [110-112]. The ketogenic diet is high in fat, ample protein and low amounts of
carbohydrates.
The mechanisms by which this diet has an effect are not clear, but a proposed
mechanism is the direct inhibitory effect of PUFAs [113]. PUFAs are vital for normal brain
development and function [114] and have been found in higher concentration in the blood
serum [115] and brain [116] following a ketogenic diet treatment. While PUFAs are
important regulators of neuronal excitability by modulating sodium and calcium currents
[117,118], less is known about their effects on KV channels. It has been reported that
epilepsy is caused by mutations in the Kv1-type and KQT-type KV channels leading to an
increase in K+ current [119-124]. The molecular mechanism of how PUFAs exert their
effects on KV channels has been the subject of debate.
D.7 PUFA mediated modulation
PUFAs have a lipophilic acyl tail with two or more double bonds and a negatively charged
carboxyl head group. Different methods of PUFA modulation have been proposed. PUFAs
binding directly to the channel or PUFAs affecting the channel by inducing changes to the
membrane properties such as fluidity have been suggested [104,105,107,125-127].
PUFAs have been suggested to electrostatically affect KV channels by binding to a
hydrophobic environment. The charge of the PUFA head group has proved necessary in the
shift of the voltage dependence of the channel. Also acyl chain properties such as the
requirement of at least two double bonds in the cis geometry are needed for the PUFA
effect on KV channels [128]. The site of PUFA modulation in KV channels has been proposed
to be close to the lipid facing side of S3 and S4 helices on the extracellular side of the
channel. Also, PUFAs are thought to act on the final voltage sensor transition and with this
effect relying on certain amino acid residues on the extracellular side of S4 [129].
Chapter D. Voltage-gated Ion Channels
24
Studies III and IV of this thesis have been dedicated to study the interactions
between PUFAs and the Shaker KV channel. I have focused on finding a PUFA binding site,
characterize the nature of these interactions, and the effects that PUFAs infer on the
energetics of the channel in going form an open to an intermediate closed state.
Chapter E. Methodology
25
Chapter E
Chapter E. Methodology
26
Methodology
E.1 Molecular modeling
Experimentally obtained structures of biological macromolecules is large and ever
increasing. The three-dimensional biological structures of all proteins and other large
molecules determined from all experimental techniques are stored online in the Research
Collaboratory for Structural Bioinformatics’ repository the Protein Data Bank (PDB). A predominant and well-established experimental technique to decipher protein
structure is X-ray crystallography. In this method, the molecule is bombarded with X-rays,
which diffract into many specific directions. From the angles and intensities of the
diffracted beams a three-dimensional description of the density of the electrons are
produced. Through this density, the positions of atoms and the chemical bonds that
connect them are determined. What are needed are well-ordered crystals and the right
crystallization conditions, which can be difficult for membrane proteins especially. A low
resolution crystallographic protein structure only indicates the overall shape of a protein,
while a higher resolution (> 2 Å) structure reveals the atomic positions of the protein. A
crystallographic structure only represents an average over all the molecules in the crystal so
there is always uncertainty of the quality of the structure due to the dynamic nature of
proteins. Another method that is widely used for structure determination is nuclear
magnetic resonance (NMR). In NMR, the magnetic spin properties of atomic nuclei are used
to set up a list of distance constraints between atoms from which the three-dimensional
structure of the protein can be resolved. This method is restricted to the determination of
the structure of smaller proteins. And as opposed to X-ray crystallography that requires
crystals, NMR structures are obtained from concentrated protein solutions [130].
Chapter E. Methodology
27
Still little or no structural information is available for many proteins. Consequently,
modeling which is the process of generating a so called ‘model’ or an idealized description of a system has provided valuable when structure determination has been beyond the
reach of experiments. Therefore methods for predicting tertiary and secondary structure
from amino acid sequence have been called for. The ab initio method attempts to solve
structure prediction from first principles, however, one needs to explore the entire
conformational space of the molecule in order to find the best possible structure. Finding a
structure with the lowest energy among an infinite number of conformations has been next
to hopeless in the present time. In search for more attainable prediction methods, a rather
common and reliable technique for modeling is comparative modeling that relies on
detecting evolutionary similarity between the sequences of an unknown and known protein
structures. Besides the comparative modeling approach, in particular with cases where low
sequence similarity hinders detection of related homologues, protein structures can be
predicted using threading [131].
E.1.1 Comparative modeling and threading
The modeling procedure requires a series of consecutive steps usually repeated iteratively
until a reasonable model is obtained: finding and selecting one or more suitable template
proteins related to the target; alignment of the target to the template sequences; building
an initial model of the target based on the three-dimensional structure of the templates; ab
initio modeling of side chains and loops in the model that differ from the template; refining
and evaluating the final model.
Comparative modeling is based on the general understanding that evolutionary
related sequences have similar three-dimensional structures. To identify a statistically
significant relationship between the target and templates a sequence alignment method is
applied. The search for related sequences is done normally by PSI-BLAST [132]. It creates a
multiple sequence alignment by an iterative process, where each time a new sequence is
identified it is included in the next query to the sequence database. It is this process which
enables the detection of distantly related sequences. A further improvement in the search
of related sequences is to prepare sequence profiles for all known structures and do a
pairwise comparison; another way to find conserved motifs among sequences has been the
implementation of profile-based Hidden Markov Models (HMM) improved with the
incorporation of predicted secondary structure elements [133,134].
When the sequence similarity between proteins is more than 70 % it is easier to
determine an alignment that leads to a reliable model. However, as this sequence similarity
decreases it becomes increasingly harder to fully trust the model and as the sequence
similarity falls below 20-30 % one enters the so called twilight zone. At this stage, threading
can be a suitable option. Threading is based on the pairwise comparison of a protein
sequence and a protein structure, where the target sequence is matched to a library of
three-dimensional profiles or threaded through a library of three-dimensional folds [135].
After finding related sequences, selecting the optimal templates is the next
important step. As a matter of fact, in selecting templates, the use of several templates can
improve the quality of the model. By combining multiple templates, different domains of
the target sequence can be aligned to the different templates with little or no overlap in
between allowing the generation of a model for the whole target sequence. Also, by
Chapter E. Methodology
28
aligning the entire target sequence to all templates, a model can be improved by local
selection of the best template.
In essence, when constructing of a three-dimensional model first the amino acid
backbone for the structurally conserved regions is generated to which loops are added
followed by the placement of side chains. This model then undergoes refinement by being
subjected to energy minimization. The procedure that Modeller [136] adopts in
constructing the model is deriving a large number of restrains. These constraints are
normally distances between alpha-carbon atoms, residue solvent accessibilities or side
chain torsion angles.
E.2 Molecular dynamics simulations
Proteins are inherently flexible molecules and can experience complex internal motions.
These motions include larger global deformations of the entire protein induced by contact
with its surrounding environment whether it is the solvent or the lipid bilayer or the folding
and unfolding of a protein or the transitioning from one state to another. But these motions
can also be local caused by internal movement triggered by constant atomic motions or
side chain fluctuations. The function of proteins can be coupled to the dynamics of these
systems, in which structural fluctuations are closely connected to the activity of the
molecule.
Simulations have played a crucial role in studying the dynamics of these
macromolecules. The properties of a system can be explained by the detailed description of
particle motions as a function of time that is obtained by molecular dynamics simulations.
Quantum mechanics methods give the most accurate results when studying the dynamics
of a system; in quantum mechanics electrons are treated explicitly and the energy of the
system is calculated by solving the Schrödinger equation, however, these methods are
limited to small scale systems usually involving a few atoms only and to much shorter time
scales, normally within picoseconds. Methods based on molecular mechanics (MM) on the
other hand permit the study of much larger systems involving millions of atoms up to
microsecond time scales normally. Molecular mechanics covers a range of simulation
techniques, but MD is one of the most common techniques applied when studying
biological macromolecules.
E.2.1 Theory
The first molecular dynamics simulation of a biologically relevant macromolecule was
carried out by McCammon et al. [137] in 1977. Performed in vacuum, the bovine pancreatic
trypsin inhibitor (BPTI) was simulated for 9.2 ps and revealed the flexible nature of proteins
as opposed to the relatively rigid structures they had been thought of. In molecular
dynamics, atomic motion is simulated by solving Newton’s equation of motion simultaneously for all atoms in the system. In classical molecular dynamics, electrons are
not treated explicitly, but rather electronic interactions are estimated and taken into
account implicitly. In molecular dynamics, forces between atoms are calculated using an
empirically derived force field, which describe the interaction between different types of
atoms using a set of mathematical functions and parameters.
Chapter E. Methodology
29
In molecular dynamics, the calculations of a system consisting of 𝑁 particles are
split into a series of very small time steps, typically between 1 femtosecond and 10
femtoseconds. At every step and for each atom 𝑖, forces on the atom with mass 𝑚 is
calculated and combined with the positions and velocities to determine new positions and
velocities by directly solving Newton’s second law of motion:
𝐹⃗ = 𝑚 ∙ 𝑎⃗ = 𝑚 ∙ 𝑑 𝑟⃗
𝑑𝑡
(E.1)
The force on atom 𝑖 is calculated by differentiating the potential energy function 𝑉 with
respect to each atom’s position:
𝐹⃗ = −∇ 𝑉(𝑟⃗ , … , 𝑟⃗ )
(E.2)
Finite difference methods are used to integrate the equations of motion. The basic idea is
that integration is divided into many small steps each separated by a fixed time. Integration
of Equation E.2 given the position of atom 𝑖 at time 𝑟⃗ (𝑡), gives the positions and velocities
at a later time step, 𝑟⃗ (𝑡 + Δ𝑡), approximated by a standard Taylor series expansion:
𝑟⃗ (𝑡 + Δ𝑡) = 𝑟⃗ (𝑡) +
𝑑𝑟⃗ (𝑡)
𝑑 𝑟⃗ (𝑡) Δ𝑡
Δ𝑡 +
+⋯
𝑑𝑡
d𝑡
2
(E.3)
There are many algorithms that exist for integrating the equations of motion using the
Taylor series expansions. A commonly used integration algorithm in molecular dynamics
simulations is the Verlet algorithm [138]. The Verlet algorithm uses the positions 𝑟⃗ (𝑡) and
accelerations 𝑎⃗ (𝑡) at time 𝑖 and the positions from the previous step 𝑟⃗ (𝑡 − Δ𝑡) to calculate
new positions:
1
𝑟⃗ (𝑡 + Δ𝑡) = 𝑟⃗ (𝑡) + 𝑣⃗ (𝑡)Δ𝑡 + 𝑎⃗ (𝑡)Δ𝑡 + ⋯
2
1
𝑟⃗ (𝑡 − Δ𝑡) = 𝑟⃗ (𝑡) − 𝑣⃗ (𝑡)Δ𝑡 + 𝑎⃗ (𝑡)Δ𝑡 − ⋯
2
(E.4)
(E.5)
The addition of the above two equations gives:
𝑟⃗ (𝑡 + Δ𝑡) = 2𝑟⃗ (𝑡) − 𝑟⃗ (𝑡 − Δ𝑡) +
𝐹⃗ (𝑡)
Δ𝑡
𝑚
(E.6)
The acceleration expressions as part of Equations E.4 and E.5 have been substituted with
the force and the mass of the particle in the additive Equation E.6 as stated by Equation
E.1. The Verlet algorithm is simple to implement, accurate and has been proven to be
stable. However, one disadvantage with the Verlet algorithm is that velocities are not
directly generated. Therefore, many variations on the Verlet algorithm have been
Chapter E. Methodology
30
developed. One such alternative is the leap-frog algorithm [139]. The leap-frog algorithm
calculates velocities at half time steps:
1
1
𝑣⃗ 𝑡 + Δ𝑡 = 𝑣⃗ 𝑡 − Δ𝑡 + 𝑎⃗ (𝑡)Δ𝑡
2
2
(E.7)
By evaluating this equation, velocities at time 𝑖 can be calculated from:
𝑣⃗ (𝑡) =
1
1
1
[𝑣⃗ 𝑡 + Δ𝑡 + 𝑣⃗ 𝑡 − Δ𝑡
2
2
2
(E.8)
The leap-frog algorithm results in greater numerical accuracy as it does not require the
calculation of subtraction of large numbers. The most challenging task in molecular
dynamics simulation is calculating the force on each particle in the system. This therefore
calls for an integration algorithm that conserves energy and momentum, is time-reversible,
and allows for a longer time step, Δ𝑡, to be used. With longer time steps, fewer iterations
are needed to sample a configurational space and obtain a trajectory of the motions of
every atom in the system. The software package used throughout this work for performing
molecular dynamics simulations in an efficient manner is GROMACS [140]. Other
commonly used packaged are NAMD [141] and CHARMM [142].
E.2.2 Force fields
At the heart of a classical molecular dynamics simulation lies the force field, which is a
description of the potential energy of a system relating the interactions between the
particles in that system. The main components of a force field consist of an energy function
together with the parameters as part of the function. Many force fields embody distinct
components in the system; the bonded interactions mediated by bonds, angles, and
torsions and non-bonded interactions related by van der Waals and electrostatics:
𝑉
𝑘 (𝑏 − 𝑏 ) +
= +
𝜀
𝑘 (𝜃 − 𝜃 ) +
𝑅
,
𝑟
−2
𝑘 (1 + cos(𝑛𝜔 + 𝛿))
𝑅
,
𝑟
𝑞𝑞
+
4𝜋𝜖 𝜖 𝑟
(E.9)
The first term in the Equation E.9 models the covalent bond interaction modeled by a
harmonic potential that gives the increase in energy as the bond length 𝑏 diverges from the
reference bond length, 𝑏 . The second term, also modeled by a harmonic potential, sums
the angles where 𝜃 is the angle formed between atoms 𝐴 − 𝐵 − 𝐶. The third term is the
energy associated with torsional rotation around the middle bond of atoms 𝐴 − 𝐵 − 𝐶 − 𝐷.
The rotational angle is determined by 𝜔 and 𝑘 sets the energy maximum. The fourth term
accounts for the non-bonded contributions in the system. Calculated for all pairs of atoms 𝑖
and 𝑗, the non-bonded interaction is calculated for atoms that are separated by at least
three bonds in the same molecule or are in different molecules. In its simple form, this term
Chapter E. Methodology
31
is modeled by a Lennard-Jones potential for the van der Waals interactions and a Coulomb
potential for the electrostatic interactions. The Lennard-Jones is characterized by an
attractive and a repulsive part. The attractive contribution is due to the dispersive force.
Also known as the London force, instantaneous dipoles form due to fluctuations in the
electron clouds and they in turn trigger the dipole formation in neighboring atoms causing
the attraction. The repulsive contribution is due to short-range repulsive forces also known
as the overlap forces that restrict electrons from occupying the same inter-nuclear region.
The energy minimum is specified by 𝜀 and is located between atoms at a distance 𝑅 .
The Coulombs potential models the electrostatic potential between particles, where 𝑞 and
𝑞 are the charges, 𝜀 is the vacuum permittivity and 𝜀 the relative permittivity of the
environment.
The summation of all these terms gives the total potential energy 𝑉. Normally high
amounts of energy are required for bonds stretching and angles bending to deviate from
their reference values. Indeed, it is energies concerned with torsional and non-bonded
interactions that mainly contribute to the divergence in structure and relative energies. The
performance of a force field relies upon the quality of its parameters. Parameterization is a
complicated process where values for the parameters are obtained by experimental data
and quantum mechanics calculations. The optimization of the parameters relating to the
torsional and non-bonded interactions is in particular important as these parameters
mainly affect the force field.
E.3 Free energy calculations
The expression free energy calculations refers to a set of simulations that calculate the free
energy difference between different states of biological systems. Calculating this free
energy difference has many applications, for example by calculating the free energy of
receptor-ligand binding, one can obtain the dissociation constant of the receptor.
Free energy by definition is a thermodynamic quantity that is equivalent to the
amount of energy accessible in a system to do work. It describes the stability of a system
since the free energy of a system is minimized only when the system is at equilibrium. The
free energy is either expressed as the Helmholtz function, 𝐴, or the Gibbs functions, 𝐺.
Systems that have a constant number of particles, temperature and volume are suitable for
the Helmholtz free energy. Most molecular dynamics experiments are however performed
under conditions of constant number of particles, temperature and pressure, making the
Gibbs free energy more suitable.
Gibbs free energy, also known as ∆𝐺, allows the prediction of the spontaneity of a
chemical reaction. If ∆𝐺 < 0, then the reaction is spontaneous, whereas if ∆𝐺 > 0, the
reaction is nonspontaneous. Gibbs free energy combines enthalpy (𝐻) and entropy (𝑆) and
is defined as:
∆𝐺 = 𝐻 − 𝑇𝑆
(E.10)
where 𝑇 is the absolute temperature. The Helmholtz function, which gives the free energy
of a system at constant number of particles, volume and temperature, is given by:
Chapter E. Methodology
32
𝐴 = −𝑘 𝑇𝑙𝑛𝑄
(E.11)
where 𝑄 is the partition function of the system and 𝑘 is Boltzmann’s constant.
Free energies cannot be accurately calculated from a molecular dynamics simulation since
these simulations do not sample the regions from phase space that play an important role
in the free energy of the system. Molecular dynamics simulations search for the lower
energy phase space and since they do not sample the high energy regions of the phase
space, calculating energies with the traditional simulation method end up in poor results of
free energies. Conventional free energy calculation methods use molecular simulations to
generate independent samples from the equilibrium distribution of the system. These
samples are consequently analyzed with statistical mechanics methods that give estimates
of free energy differences.
E.3.1 Sampling along a reaction coordinate
When studying the behavior of a system in specific states or along particular paths, one is
often interested in calculating the free energy along a reaction coordinate. The reaction
coordinate could be the distance between two molecules, a torsion angle along the protein
backbone, or the relative orientation of a helix with respect to the membrane. The free
energy or potential of mean force (PMF) [143] along a reaction coordinate is given by:
𝐴(𝜉) = −𝑘 𝑇𝑙𝑛𝑃(𝜉)
(E.12)
where 𝑃(𝜉) is the probability distribution of the reaction coordinate 𝜉. To sample important
values of the reaction coordinate, methods such as umbrella sampling [144] can be used to
access regions of low probability in the conformation space. In umbrella sampling, the
interval along the reaction coordinate is split into subintervals, called windows, in which a
biasing potential can be used to improve sampling.
E.3.2 Accelerated weight histogram
A drawback with umbrella sampling for example is that this method only samples one
pathway at a time except if the sampling is carried out multiple times. In a transition,
however, there could be many different pathways involved and as such, efficient sampling
is vital in mapping out an accurate free energy landscape. Here is where a recently
developed approach, the accelerated weight histogram (AWH) method [145] is
advantageous. The AWH method is an extended ensemble method that uses an adaptively
biasing approach. A simulation in extended ensemble performs a random walk in the
energy space and overcomes the difficulty of getting stuck in local energy minima. By using
a Gibbs sampler and a probability weight histogram the AWH method enables larger
transitions in parameter space. The use of the weight histogram allows for efficient
utilization of the transition history in order parameter space in the free energy updates by
Chapter E. Methodology
33
adapting the bias as well in keeping a track of the simulation. Also, by using a weight
histogram the need of discretization of the parameter space is removed. AWH estimates
the free energy by:
𝑔(𝜆) = 𝑓(𝜆) + ln 𝜌(𝜆)
(E.13)
where 𝑔(𝜆) is the biasing function, 𝑓(𝜆) the estimate of the true free energy and 𝜌(𝜆) the
target distribution. During the simulation, the biasing function is adjusted to fit a specified
target distribution. To begin with, 𝑓(𝜆) is initialized but throughout the sampling it is
iteratively refined. The main benefit of the AWH method lies with the iteration of the
process of sampling along the reaction coordinate, permitting the sampling of different
pathways.
Chapter F. Study I
34
Chapter F
Chapter F. Study I
35
Study I
F.1 Structural modeling of the N-terminal signal–
receiving domain of IB
A detailed understanding of molecular mechanism of NF-κB activation, regulation and the
protein-protein communication with partners may assist to design and develop novel
chronic inflammation modulators as well as anti-cancer drugs. The insight gained from
structural biology of NF-κB and IκBα proteins and its implications for the signaling process
control have been reviewed extensively by i.e. [146-148].
From the structure of IκBα in complex with NF-κB, a valuable level of insight was
rendered into the regulation of NF-κB by IκBα and the nature of their association [61,149].
Each unit of the complex was partially truncated leading to a missing IκBα N-terminal
segment comprising ~70 residues. This N-terminal SRD receives the phosphorylation and
ubiquitination signals and targets the protein to the proteasome for degradation [150], but
has no measureable effect on binding of IκBα to NF-κB [151]. While SRD plays a crucial part
in activation of NF-κB, it has not been found to be engaged in enabling the complex
formation of IκBα/NF-κB [152-154]. Protein crystallization and structure determination
were unsuccessful for free IκBα due to its short lifetime and degradation within minutes.
This led to the suggestion of conformational disorder in the free protein [65]. For IκBα in
complex with NF-κB, however, there are two protein crystal structures available (PDB IDs
1IKN and 1NFI). The truncated IκBα sequences in 1IKN (residues 73-292) [61] and 1NFI
(residues 71-280) [149], however, did not reveal information about possible secondary
structure elements (SSEs) in the SRD.
Chapter F. Study I
36
The concept of conservation of secondary structure elements [155] in families can
be used to identify proteins only distantly related in sequence, which may, however, still
share a higher degree of conservation of SSEs. Recent approaches have demonstrated that
the use of multiple tools of secondary structure prediction and the use of a ‘consensus’ of methods yields more reliable results than single algorithms [59,60].
In this study, I present for the first time, the structural elements of the full length
SRD of IκBα in complex with NF-κB and in free IκBα. I clearly show that the SRD displays
well-defined secondary structure elements and cannot be considered as ‘unstructured’. In contrast, it contains 3 α-helical regions which are stable during molecular dynamics
simulations. Also, in free IκBα the SRD is structured albeit displaying a larger degree of
flexibility and larger fluctuations. This represents the first step in an approach to model the
signal transduction cascade of the NF-κB/IκBα complex from IKK phosphorylation to
degradation.
F.2 Methods
F.2.1 Structural modeling
The secondary structure prediction of the full IκBα sequence was performed with
SYMPRED [156] which builds upon results from PROF [157], SSPRO [158], YASPIN [159]
and PSIPRED [160]. In addition, JPRED3 [161], JUFO [161], NetSurfP [162], PORTER [163],
PredictProtein [164] and ScratchProteinPredictor [165] were also used. All secondary
structure prediction algorithms correctly identified and positioned the six ankyrin repeat
units in the crystal structure (PDB ID 1IKN) in addition to four additional α-helical regions in
the N-terminal SRD, which is not resolved in the protein crystal structure. The consensus of
predicted secondary structure elements was used for structural modeling of the SRD.
To identify a suitable structural template for modeling the SRD of IκBα,
pDomThreader [166] was used; a profile based recognition fold method incorporating
domain superfamily discrimination, which distinguished 46 probable structural templates.
The fourth ranked template, 1N11, was chosen as a suitable template basing the decision
on a top alignment score, the degree of coverage and the structural alignment of 286 out of
317 residues in IκBα. pDomTHREADER [166] makes use of the CATH database of
annotation of protein structural superfamilies from the PDB [167]. It is an implementation
of GenTHREADER, a method which predicts protein fold from sequence by integrating
profile-profile alignments, secondary structure gap penalties and both classic pair and
solvation potentials employing an optimized regression SVM model. pDomTHREADER is
thus able to discriminate between different structural superfamilies from the protein
sequences and to detect distant homology to proteins of known structure.
The structural model was generated with Prime [168,169]. A manually constructed
sequence alignment of our templates 1IKN and 1N11 was used. A two-template composite
model was thereby constructed; residues 73-292 were based on the crystallized IκBα
protein 1IKN while the secondary structure elements of ankyrin protein 1N11 served as
basis for residues 1-98. The Build process involves coordination of the copying of the
backbone atoms for aligned regions and side chains of conserved residues, building
Chapter F. Study I
37
insertions and deletions in the alignment, optimization of side chains not found in template
and energy minimization of those residues not derived from the templates. The Prime Build
process applies the OPLS_2005 all-atom force field for energy scoring and the Surface
Generalized Born (SGB) continuum solvation model for treating solvation energies and
effects. Additionaly it utilizes the residue-specific side-chain rotamer and backbone
dihedral libraries, derived from the non-redundant data sets extracted from the PDB.
F.2.2 System assembly and protocol for MD simulations
The MD simulations were carried out using Gromacs 4.5 [170,171] employing the
GROMOS96 43a1 force field [172]. The all-atom structural model of IκBα bound to the X-ray
crystallographic structure of NF-κB included 6945 atoms in total. The protein complex was
immersed in a rectangular box of dimensions 78 x 89 x 145 Å 3 solvated with 29686 SPC
water molecules together with 117 Na+ and 91 Cl- ions in order to neutralize the net system
charge. The structural model of the free IκBα was immersed in a slightly smaller rectangular
box of dimensions 66 x 68 x 114 Å3 solvated with 15757 SPC water molecules together with
70 Na+ and 47 Cl- ions, in total containing 50270 atoms. The LINCS algorithm [173] was
applied for constraining bond lengths. Electrostatic interactions were calculated every step
with the Particle-Mesh Ewald algorithm [174]. Neighbor lists were saved and reused for 5
steps. The simulations were performed at constant pressure of 1.0 bar with ParrinelloRahman pressure coupling and the isotropic pressure scaling, time constant of 1.0 ps, and a
system compressibility of 4.5e-5 bar-1. The temperature of the system was coupled to 300 K
using the velocity-rescaling algorithm with a time constant of 0.1 ps. Newton’s equations of motion were integrated using the leap-frog algorithm with a 2 fs time step.
The solvated system was first minimized with the steepest descent algorithm until a
maximum force of < 100.0 kJ/mol was reached. Equilibration of the system was initiated by
10000 steps of position-restrained MD by relaxing the solvent and keeping the nonhydrogen atoms of the system fixed. With the system relatively free of strain an NVT
equilibration phase followed by an NPT phase of 10000 steps each was then carried out.
Coordinates were saved every 2 ps for analysis and the production phase of the simulation
ran for a total of 200 ns. Fan and Mark have shown that molecular dynamics simulation in
explicit water are able to refine homology-based protein structures within a short period of
simulation [175]. For small to medium-sized proteins (50-100 amino acids), the first 1-5 ns
were able to remove initial distortions and only in few cases simulations of > 100 ns were
necessary to obtain a significant reduction of RMSD. This was taken as a lower threshold
and added a factor of two considering the complexity of the system. Three independent
replicates of the system were simulated for 600 ns in total, each starting with different
initial velocities. Simulating independent replicates is a rather cost-effective way to sample
conformational space [176].
F.3 Results and Discussion
F.3.1 Structural elements in the signal-receiving domain (SRD)
Chapter F. Study I
38
In order to better understand the effects of phosphorylation and the mechanisms, which
govern recognition of phosphorylated IκBα, and consequentially initiate ubiquitination, one
requires a structural model of the complete N-terminal protein SRD. A BLASTp search of
the first 72 residues of IκBα did not yield any significant sequence similarity with other
known proteins.
A consensus-based secondary structure annotation of the full length IκBα sequence
with SYMPRED [156] was performed which builds upon results from PROF [157], SSPRO
[158], YASPIN [159] and PSIPRED [160] (Figure I.1).
Figure I.1 Consensus secondary structure annotation of full length IκBα (residues 1–317).
Truncated, as crystallized IκBα from 1IKN (residues 73–293), is shown in bold letters. The six
ankyrin repeat units of the ARD are recovered, correctly annotated and positioned. Two
additional α-helix-loop-α-helix regions were detected in the N-terminal SRD. The Cterminal PEST domain displays less structural features.
In addition, JPRED3 [161], JUFO [161], NetSurfP [162], PORTER [163], PredictProtein [164]
and ScratchProteinPredictor [165] were also employed and give close to identical results.
All six ankyrin repeat units in the crystal structure (PDB ID: 1IKN [61]) are recovered,
correctly annotated and positioned. In addition four α-helical regions were detected in the
N-terminal SRD, which is not resolved in the protein crystal structure (Figure I.1). This
indicates that the SRD region may contain secondary structured subregions with a high αhelical content (residues 11-14, 21-29, 44-50, 56-62) not covered in any of the available IκBα
crystal structures and not investigated in any of the NMR studies of free or complexed IκBα.
The position of these α-helices is not fixed with respect to each other and may obstruct
protein crystallization of full length IκBα.
The initial finding prompted the generation of a full-length IκBα model including the
SRD and the investigation of its spatial and temporal integrity and stability. Due to the
absence of sequence similarity of the SRD region to any structurally resolved protein in the
PDB (~12 %) sequence-based comparative modeling is not a feasible approach here. As an
alternative, the choice of template was based on identification of a remote structurally
Chapter F. Study I
39
related protein template with a similar secondary structure fold. The conservation of SSEs
in protein superfamilies can guide the design of a structural model. Even when the structure
of only a single member of a superfamily is known the conservation of SSEs can be used to
predict the structure of other superfamily members [177,178]. Such information is useful
when modeling the structure of other members of a superfamily or identifying structurally
and functionally important positions in the fold. An efficient template detection allows the
structural modeling to be extended even in the twilight zone of 10–30 % sequence identity
[178].
pDomTHREADER [166] identifies 46 possible structural templates with reliable
secondary structural similarity. Based on a top alignment score, the degree of coverage and
the structural alignment of 286 out of 317 residues in IκBα, one of the top ranked structures
(1N11) was chosen as a template for modeling IκBα.
As an alternative approach, a combination between comparative modeling and de
novo protein structure prediction was performed using Robetta. For proteins with detected
PDB homologs, comparative models are built based on templates that are found and
aligned with incorporated versions of HHSEARCH/HHpred, RaptorX, and Sparks-X. Protein
domains with no close PDB homologs are generated with the Rosetta de novo protocol
[179,180]. A structure prediction carried out by Robetta [181] for the full IκBα sequence also
yielded 1N11 as the top-ranked template of choice for the generation of its structural
models. Figure I.2 shows the alignment of secondary structure elements of IκBα and 1N11
in the SRD region.
Figure I.2 Two-template sequence alignment used for the generation of a composite
structural model of the full-sequence IκBα. The bold segments in each template correspond
to the α-helical regions forming the ankyrin repeat units present in the crystal structures of
1IKN and 1N11. The curved boxes in red display the helical segments in our generated
structural model of IκBα.
Despite an overall low primary sequence identity of only 23 %, the alignment of
secondary structural elements is striking. 1N11 is the crystal structure of a 12 ankyrin repeat
units stack from the human ankyrinR. AnkyrinR belongs to a family of adaptor proteins that
Chapter F. Study I
40
mediate anchoring between integral membrane proteins and the spectrin-actin
cytoskeleton. The membrane-binding domain of ankyrins contains 24 ankyrin repeats of
which the crystal structure of the human ankyrinR maps the D34 region. This region, which
consists of repeats 13-24, is stacked contiguously in the shape of a left-handed superhelix
[182].
A composite model from crystallized IκBα (67-317) 1IKN and ankyrinR 1N11 PDB
structures was generated. Residues at positions 73-292 were taken from the crystallized
IκBα protein (PDB ID: 1IKN) and for residues 1-98 SSEs of the SRD were taken from the Xray structure 1N11. For an overlapping stretch of residues 73-98, two α-helices forming one
ankyrin repeat in the 1N11 template was taken to remove any possible artifacts from
truncated sequence crystallization.
F.3.2 Structural refinement by molecular dynamics simulations
The protein-protein complex model was used as a starting configuration for subsequent
MD refinement. The stability of the suggested secondary structural elements in the SRD
and the dynamics of possible rearrangements were investigated.
In order to achieve a reliable full-sequence structural model, three independent MD
simulations of IκBα in complex with NF-κB were performed for 200 ns each in a neutralized
solvent box of about 30000 explicit water molecules. Thus, a total production simulation
time of 600 ns was achieved. After energy minimization, a stepwise relaxation of the
simulation setup and careful equilibration first in an NVT and then in an NPT ensemble, the
general behavior of all simulation runs reveals well-behaved and stable systems. This is
reflected in the conservation of total energy and temperature of the entire system, which is
kept at a constant room temperature of 300 K throughout the whole 200 ns simulation
runs.
The structural stability of the IκBα/NF-κB complex is also monitored by calculating
the root mean square displacement (RMSD) from the starting protein-protein complex
structure. The RMSD increases sharply to 3.5-4.5 Å for the three replicate runs during the
first 100 ns of the simulations, and settles at roughly 4.5-5.5 Å for the last 100 ns, indicating
a well-structured stable complex. The results discussed herewith are the average findings
of the three replicate runs unless otherwise stated.
In order to investigate the secondary structural profile of the initial IκBα model and
possible structural re-arrangements, the 200 ns simulation was sectioned into two equal
parts. This provides a comparison of results at the beginning and end of the production run
periods.
To better understand the inherent flexibility of our protein, the root mean square
fluctuations (RMSF) of the backbone Cα atoms of IκBα around the average structure were
calculated (Figure I.3 A). The SRD N-terminal segment comprising ~75 residues clearly
stands out as the most flexible region, particularly in the initial 100 ns of the simulations.
Although not as expressed as in the initial 100 ns of simulations, the flexibility in the
subsequent 100 ns run region is still comparatively high. One encounters the two most
flexible helical regions of the whole protein, namely helices 1 and 2, also in this region. This
result indeed explains the difficulty to crystallize the SRD region. Instead an N-terminal
truncation of IκBα was necessary to obtain protein single crystals [65]. One sees, in general,
the retention of all crystallographically resolved six ankyrin repeat units in IκBα during the
entire simulation runs (Figure I.3 A). While the peaks mark the hairpin loop segments
Chapter F. Study I
41
connecting the α-helices in each ankyrin repeat unit, the troughs of the RMSF plot
correspond to helical regions. This result shows that while the helical regions are stable and
not so flexible, greater flexibility is observed in the β-loop segments. This is in agreement
with the amide 1H/2H exchange experiments followed by MALDI-TOF mass spectrometry
(MS) in bound and free IκBα [63]. The β-hairpins of some ankyrin repeats readily exchange
amide protons for deuterons (1st, 5th and 6th ankyrin units) whereas other units (2 – 4) are
less solvent accessible. In particular, ankyrin repeat unit 1 remains highly solvent accessible
even in the complex. The solvent accessibility of the β-hairpin in ankyrin repeat unit 1 (AR 1)
decreases slightly upon NF-κB binding [64].
Figure I.3 (A) Average root mean square fluctuations (RMSF) of the backbone of IκBα for
the initial and final 100 ns of the simulation. Shaded areas depict α-helical regions at the
end of the three independent 200 ns simulation periods. (B) Probability distribution of αhelix formation of the first 70 residues of the SRD of IκBα in complex with NF-κB.
Figure I.3 B gives the probability distributions of helical formations in the SRD of
IκBα. Together with the RMSF of Figure I.3 A, a consistent picture of stable vs. flexible
subregions in the signal receiving domain is obtained. Residues 31-37 in the SRD
immediately adjacent to the second α-helix in the N-terminal region represent the most
Chapter F. Study I
42
flexible part of IκBα, in the case of disregarding residues beyond 275 (Figure I.3 A). It is
natural to discard residues beyond 275 from the comparison as they form a long loop and
constitute a rather disordered region void of any tertiary structure. We do see the
conservation of three α-helices, residues between 8-15, 22-30 and 44-50 within the SRD
region. These values are in agreement with the predicted secondary structure models,
which identified the three α-helices to lie between residues 10-13, 22-29 and 44-50. The last
two α-helices align perfectly, while SYMPRED predicts a somewhat shorter α-helix
compared to that observed in the refined structure. Furthermore, the fourth α-helical
element, which was positioned from residues 54-63 from the 1N11 template, no longer
adopts an α-helical shape but acquires instead a less ordered loop conformation (Figure I.3
B and Figure I.4, below). Here, obviously the refinement by MD simulations is sufficient to
remove the ambiguous assignment of secondary structure elements and provide a more
stable conformation of this stretch of 10 amino acid residues in length. All other secondary
structure elements are retained during the MD simulations. This gives the confidence in the
reliability of the protein-protein complex model and the existence of well-defined
secondary structural elements in the SRD of IκBα when it is in complex with NF-κB.
Figure I.4 The secondary structure elements of the first 70 N-terminal residues of IκBα in
complex with NF-κB as calculated by DSSP for the three system replicas during the initial
100ns (A) and final 100 ns (B) of the simulation.
The time-evolution of secondary structure elements in the N-terminal SRD during
the MD refinement is then analyzed in detail. The DSSP-annotated SSEs of the first 70
amino acid residues in IκBα for each of the replica systems is plotted in order to analyze the
SSEs of the first 70 amino acid residues during the MD trajectory frames (Figure I.4). The
Chapter F. Study I
43
first three α-helices, residues between 8-15, 22-30, and 44-50 retain their α-helical structure
(blue regions) during the initial 100 ns MD simulations in all three systems replicas. (Figure
I.4 A). They are followed by a recurring β-sheet turn β-sheet formation (green-yellowgreen). This region is followed by an unstable α-helix that is formed between residues 5262. This short helix is observed only in two of the replicas (top and bottom). This segment
mainly adopts the turn/bend secondary structure in the third replica. The structural stability
is observed for the first three α-helices throughout the entire simulations during the final
100 ns of the simulation runs. (Figure I.4 B) The temporarily formed fourth α-helix,
however, observed in the first 100 ns, is no longer formed and the sequence instead
remains variable in its secondary structure. During most of the production runs, it takes a
turn-like secondary structure (yellow) or bend (green) with short interludes of stretches of
310-helices (grey) and -helices (purple).
Figure I.5 summarizes the results from secondary structure prediction, initial model
generation and secondary structure elements of the full-length IκBα obtained after MD
refinement. Four helical stretches were detected from consensus SSE prediction and thus
also represented the starting SRD model (top line, up to residue 70). After MD refinement,
three helical stretches are structurally retained and the fourth one was not stable and
adopts a disordered conformation, after the MD simulations. The ankyrin repeats of the
ankyrin repeat domain (ARD) are structurally stable during the MD simulations of the
protein-protein complex and well positioned with respect to the crystal structure.
Chapter F. Study I
44
Figure I.5 A graphical map of the secondary structure elements of IκBα, displayed on its
complete sequence. The boxes highlight the α-helical regions, and the arrows indicate βstrands. Dark green designates secondary structures determined in the crystal structure
1IKN, blue denotes secondary structures predicted by SYMPRED, and brown and
fluorescent green indicate secondary structures suggested by our initial and refined
structural models.
In Figure I.6, I present the refined structural model of the IκBα-NF-κB complex (blue)
portrayed together with the initial structural model (purple). The refined representative
structure is depicting the last frame of a system replicate that has the lowest RMSD with
respect to the average structure. This model reveals three helical structures in the
previously not resolved SRD unit in addition to the six ankyrin repeats in the ANK protein
domain. While the inner helix is nine residues long and extends from positions 22-30, the
initial helix in the first pair of helices is eight residues long, spanning from positions 8-15 in
the IκBα. The α-helix pair is followed by a 13-residue long loop, joining this element with the
Chapter F. Study I
45
consecutive α-helix of seven residues long covering positions 44-50. The lengthy loop
linking the third helix to the subsequent ankyrin repeat domain comprises 26 residues, and
connects the unresolved N-terminal segment of IκBα to the crystallized ankyrin repeat
domain of this protein. The structural superpositioning of the initial and refined models of
IκBα bound to its partner, NF-κB, reveals an ANK domain that is partly rigid and wellstructured. Ankyrin repeats 4-6 remained intact and display greater stability when bound to
NF-κB, while ankyrin repeats 1-3 show increased flexibility. This is in agreement with the
analysis of residual dipolar coupling (RDC) of free and bound IκBα which showed that the
helix 2 from ankyrin repeat 3 differed most in the free and bound forms [65].
Figure I.6 Ribbon diagrams of the three-dimensional initial structure (purple) and the
refined structure after a 200 ns MD simulation (blue) of IκBα. The structures are shown in
comparison by superpositioning IκBα’s binding partner NF-κB (gray).
In particular, ankyrin repeat 1 shows the greatest displacement, which together with
the SRD segment move away from NF-κB and deviate the most from the initial structure.
This is in agreement with experimental studies which could show that the SRD does not
contribute to the overall NF-κB binding affinity to IκBα [183]. Also, NMR studies of IκBα in
complex with its binding partner, NF-κB, show a more flexible ankyrin 1-4 domain in
comparison to rather rigid ankyrin repeats 5-6 [184]. An earlier amide H/D exchange study
[64] indicated that when in complex with NF-κB, ankyrin repeats 5 and 6 fold into compact
domains upon binding to NF-κB. Along with ankyrin repeats 5 and 6, ankyrin repeat 1 is
another region seen to display greater conformational flexibility as observed here in the
refined structure of IκBα.
F.3.3 Conformational change induced in IκBα in its bound form
to NF-κB
Chapter F. Study I
46
Thus, so far I have looked at the structural elements in IκBα only. In the crystal structures
and in the simulations however, IκBα is in complex with NF-κB (the RelA/p50 heterodimer)
and for this reason it is imperative to look at the conformation of IκBα in relation to its
binding partner, NF-κB, and see how the nature of this association was affected. The
protein-protein surface area of interaction is larger than 4000 Å2 and all six ankyrin repeat
units are involved in forming a noncontiguous contact surface. Here, I discuss in particular
electrostatic and hydrogen bonding interactions between IκBα and RelA/p50. The
hydrogen bonds that are discussed here remain intact for longer than 10 % of the
simulation time and occur in at least two of the replicate simulations.
The IκBα/RelA interface
IκBα binds to RelA by forming a number of hydrogen bonds between different regions of
each protein (Table I.1). Several residues situated on ankyrin repeats 5 and 6 form hydrogen
bonds with residues located on both the RelA dimerization subunit and the RelA aminoterminal. The IκBα carboxy-terminal residues are in close contact with regions on the
amino-terminal and dimerization subunit of RelA and form several hydrogen bonds.
Table I.1 Hydrogen bond contacts between IκBα and the p50/RelA subunits of NF-κB.
Hydrogen bonds
IκBα
GLY155
LEU157
ASN216
ASP226
THR247
GLN249
TRP258
GLY259
GLN266
GLN267
GLN271
GLN271
LEU280
SER283
ASP285
GLU286
GLU286
SER288
RelA
ARG297
ARG297
ASP243
SER238
ASP243
ASP243
GLN26
GLN241
ILE24
GLU22
LEU179
VAL21
GLN29
GLU222
GLN247
GLN247
THR191
GLN247
IκBα
ASP73
GLN107
ASN109
ASN109
TYR181
ASN182
GLY183
CYS215
TYR248
TYR248
GLN249
p50
LYS352
LYS352
ASP353
LYS352
THR256
THR256
THR256
MET253
LYS343
GLU341
ARG252
RelA
ARG198
ASN200
ASP243
HIS245
p50
HIS304
ASP254
ARG252
CYS270
Residues found in interactions in the crystal structure 1IKN are shown in bold. All Bonds are present for
more than 10 % of the total simulation time in at least 2 of the replicate simulations.
The other major source of stabilization is via electrostatic interactions from the salt
bridge interactions between the carboxy-terminal of IκBα and different regions of RelA
Chapter F. Study I
47
(Table I.2). The ARD region of IκBα contributes to the IκBα/RelA stabilization by forming
salt bridges between ASP226 and ARG218 on ANK5 and between ARG253 and ASP243 as
well as GLU211 on the dimerization component. In addition, ARG264 on ANK6 interacts
with GLU22 on the amino-terminal of RelA. Specifically the interaction between ARG218
and ASP243 is also observed to form in the crystal structure of IκBα [61].
Table I.2 Salt bridge formations between IκBα and the p50/RelA subunits of NF-κB.
Salt bridges
IκBα
RelA
IκBα
p50
RelA
p50
GLU85
GLU85
GLU86
GLU125
ARG143
GLU153
GLU153
ARG218
ARG218
ASP226
ARG264
GLU282
GLU282
GLU282
GLU284
GLU284
ASP285
GLU286
GLU287
GLU287
ASP290
GLU292
ARG302
LYS301
ARG302
ARG302
ASP294
ARG295
ARG297
ASP243
GLU211
ARG253
GLU22
ARG30
ARG158
LYS79
ARG246
LYS79
LYS218
LYS218
ARG246
LYS218
LYS221
ARG246
GLU41
GLU72
GLU72
ASP73
GLU138
ARG143
GLU286
GLU286
GLU287
ASP290
GLU292
GLU292
LYS354
LYS352
LYS354
LYS354
LYS323
GLU350
ARG305
LYS272
LYS272
LYS272
LYS249
LYS272
ARG198
ARG198
ARG201
ARG201
GLU211
ASP217
ASP243
ARG246
ASP302
GLU265
ASP254
GLU265
ARG252
ARG305
ARG252
ASP271
Residues found in interactions in the crystal structure 1IKN are shown in bold. All salt bridges occur in at
least 2 of the replicate simulations throughout the whole simulation.
The elongated and relatively flexible 13 residue carboxy-terminal of RelA, known as
the NLS polypeptide, extends across ankyrin repeats 1-3 and makes several contacts with
residues present on the loops and helical regions of these ankyrin repeats, forming both
hydrogen bonds and salt bridges.
The IκBα/p50 interface
A number of residues on ankyrin repeats 4-6 interact with the dimerization domain on p50
by forming hydrogen bonds. Among these interactions, TYR181 has previously been shown
to be a key player in the interaction between NF-κB and IκBα [61]. Eminently, residues
Chapter F. Study I
48
CYS215, TYR248, and ARG252 on the p50 subunit are among those reported to form
interactions in the crystal structure of IκBα. The amino acid residues LYS352-ASP353
located on the carboxy-terminal of p50 engage in additional hydrogen bond interaction
with the residues ASP73, GLN107, and ASN109 situated on ankyrin repeats 1 and 2. The
interaction between IκBα and p50 is further stabilized by electrostatic interactions. The
carboxy-terminal PEST sequence residues GLU286-GLU287, ASP290, GLU292 in IκBα take
part in forming salt bridges with the residues LYS249, LYS272, ARG305 on the aminoterminal and the interconnecting loops on the ‘top’ of the p50 subunit. Ankyrin repeats 1-3
and the SRD in IκBα and the carboxy-terminal and an interconnecting loop at the ‘bottom’ of p50 participate in another set of salt bridge network involving residues GLU41, GLU72ASP73, GLU138 and LYS323, LYS352, LYS354, on respective chain. Notably, with one single
exception, the acidic residues are contributed by IκBα, whereas the basic residues are to be
found on the p50 subunit.
The RelA/p50 interface
The dimerization interface takes part in several hydrogen bonds formed by 8 residues
including an ASP254(p50)/ASN200(RelA) hydrogen bond. This hydrogen bond can also be
found in the crystal structure and is considered one of the most critical interactions in
discriminating subunit dimerization specificity among NF-κB dimers [61,185,186]. The
other hydrogen bonds include HIS304(p50)/ARG198(RelA), ARG252(p50)/ASP243(RelA),
CYS270(p50)/HIS245 (RelA). The RelA/p50 dimer interface is additionally stabilized by
electrostatic interactions. Several residues form salt bridges between the two subunits.
Two of these include salt bridges that are also reported for the crystal structure namely
residues ASP217 and ASP271 on the p50 subunit and ARG305 and ARG246 on the RelA
component, respectively [61].
F.3.4 Free IκBα vs bound IκBα
All the simulations discussed above were describing the stable, long-living complex of IκBα
with its binding partner NF-κB, as revealed in their crystal structures. All efforts to
crystallize IκBα in its unbound state have been unsuccessful [63]. For this reason, additional
simulations of full-length free IκBα in solution were performed and compared with the
more stable NF-κB-bound state.
The simulation setups followed the same procedure as for the bound IκBα, resulting
in three replicated systems of 200 ns each. Conservation of total energy and temperature
of the three simulations points to systems that have reached a stable state. In contrast to
the bound IκBα, the RMSF of the free state of IκBα remains on average ~1 Å higher
compared to its complexed state. This points to a higher degree of flexibility of free IκBα
compared to its complexed state. The RMSF of the backbone of the protein around the
average structure in the modeled SRD remains the most flexible domain throughout the
protein in addition to the unstructured C-terminal region (Figure I.7 A). The probability
distributions of the helical propensity in the SRD of the bound IkBa reveal (Figure I.7 B) the
first three helical segments to be stable throughout the whole simulation. In free IκBα,
although the first three helical segments are present in all three replicate simulations, we
Chapter F. Study I
49
observe different probabilities across the different replicate simulations. Bound IκBα
displays a narrower distribution of probabilities of helical regions and this indicates to a
stabilization of the SRD upon complexation with NF-κB. The fourth initially assigned helix
in the SRD varies in both length and probability in both the bound and free forms of IκBα,
indicating that this fourth helix is not well defined and not stable during MD refinement.
Figure I.7 (A) Average root mean square fluctuations (RMSF) of the backbone of the free
IκBα (cyan) in comparison to the one in complex with NF-κB (black). (B) Probability
distributions of α-helix formation of the first 70 residues of the SRD. Left: free IκBα. Right:
IκBα in complex with NF-κB.
The secondary structure evolution of the first 70 amino acid residues in the SRD of
the free IκBα (Figure I.8 A) reveals greater differences in the SRD in terms of secondary
structure element evolution in comparison to the bound IκBα. The first helix in the free IκBα
is considerably shorter than its counterpart in the bound IκBα. During the first 100 ns of the
simulations, this helix can be clearly distinguished whereas it is only present in 2 of the
replicate runs in the final simulation period. The second and third helices remain intact
throughout the entire 200 ns simulations in all three replicate runs, which is very similar to
the pattern seen in the bound IκBα simulations. In contrast to the bound IκBα, here one
observes the formation of a 4 residue long fourth helix in 2 of the replicate runs; in one of
the simulations this helix is present during the entire simulation, whereas in the other run it
appears in the last 100 ns of the simulations with irregular intervals. In a previous study
[68], the conformations of a short 24 amino acid peptide (residues 21-44) of the doubly
phosphorylated free IκBα were characterized by NMR spectroscopy and MD simulations
and compared to its β-TrCP bound state using saturation transfer difference NMR. The
Chapter F. Study I
50
conformational observation agreed on the presence of a bend between residues 30 and 36
in both states of the phosphorylated peptide, a trend which I also observe throughout the
simulations of the free and NF-κB bound states of IκBα. While the N-terminal of amino
acids 30 to 36 is preceded by a short α-helix and the C-terminal succeeded by a region of βsheet–turn–β-sheet flanked by bends in the free and bound states of IκBα in this study,
Pons et al. observed disordered N- and C-terminal segments in the free IκBα versus the
adoption of turns in the bound state IκBα. This difference in results can be rationalized from
the truncation of the peptide which could have influenced the conformational integrity of
the N- and C-terminals, an effect which would not be detectable in our structural models of
the full-length IκBα.
Figure I.8 (A) The secondary structure elements of the first 70 N-terminal residues of free
IκBα calculated by DSSP for the three system replicas for the entire simulation. (B)
Interatomic distance matrices for the first 70 N-terminal residues of free IκBα (top) and in
complex with NF-κB (bottom).
Figure I.8 B shows the interatomic distance matrices depicting the smallest distance
between residue pairs in the SRD of IkBa for both free (top) and complexed IκBα (bottom).
The distance matrices of all three replicates are very similar and there are no large
differences in interatomic distances upon NF-κB binding. The red and yellow colors indicate
shorter distances between the residues and are more detectable for regions where helical
segments are present in the SRD. In both the unbound and free forms of IκBα, the fourth
segment is less apparent across the replicates.
There are, however, also apparent stretches of amino acids which display a higher
degree of flexibility upon NF-κB complexation (Figure I.7 A). The residues around positions
133 and 167 become more flexible upon protein-protein complex formation. These
positions correspond to loop regions following the outer helices in AR2 and AR3. This was
also found by analyzing residual dipolar coupling (RDC) of ARs 1-4 [65].
Another interesting comparison between the free and bound IκBα structures is the
solvent accessible surface area (SASA) or the relative solvent accessible area (RSA) of the
Chapter F. Study I
51
phosphorylation and ubiquitination sites located on the SRD (Table I.3). These sites (SER32
and SER36 for phosphorylation and LYS21 or LYS22 for ubiquitination) ought to become
accessible by the kinase IKK and the E3 ligase, respectively, in the complexed form of IκBα.
The RSA is computed by the SASA of the residue normalized by the accessible surface area
of that residue in its extended tri-peptide (Gly-X-Gly) conformation. By setting a threshold
of < 20 % for buried residues, SER32 and SER36 are both surface-exposed in the bound
IκBα, while in its free state only SER32 lies above the threshold. SER36 in the free state has
an RSA of 9.3 %, which is considerably lower than the threshold and can be considered to
be a buried residue. As regards to the ubiquitination sites, LYS21 stays well buried in both
the free and bound states of IκBα. However, LYS22 with an RSA of well over 60 % in both
states of IκBα remains surface exposed. Thus, in the bound-form of IκBα the
phosphorylation SER32 and SER36 sites are accessible by the IKK and we suggest LYS22 to
be the putative site of ubiquitination.
Table I.3 Solvent accessible surface area (SASA) and relative surface area (RSA) of the
free and bound IκBα.
SASA (Å)
RSA (%)
Bound IκBα
Free IκBα
Bound IκBα
Free IκBα
SER32
61.2±18.5
43.4±15.7
50.2±15.2
35.6±12.8
SER36
50.6±14.0
11.3±9.1
41.5±11.5
9.3±7.4
LYS21
18.5±4.8
39.2±16.4
8.8±2.3
18.6±7.8
LYS22
135.6±6.0
127.5±17.9
64.3±2.8
60.4±8.5
The accessible surface areas of serine and lysine are 122 Å and 211 Å respectively as calculated by Miller
et al. [187]. The SASA values are the averages of the three replicate simulations over the 200 ns of total
simulation time.
Chapter G. Study II
52
Chapter G
Chapter G. Study II
53
Study II
G.1 Double phosphorylation-induced structural changes
in the signal-receiving domain of IB in complex with
NF-κB
The findings in study I provide the starting platform for investigating and understanding
the mechanism of phosphorylation in IB by IKK and the effect of single and double
phosphorylation by IKK. In an approach to unravel the molecular basis of NF-κB signaling
by multiscale molecular simulations, the structural model of full length IB obtained in
Study I was further refined for this study (Figure II.1 A).
Mutation of either serine residue at sites Ser32 and Ser36 (Figure II.1 C) have been
shown to disrupt polyubiquitination of IB and abolish the stimulus-induced degradation
of IB, thereby requiring the strict phosphorylation of both serine residues as a
recognition and initial step in NF-B activation [47,188-192]. Although phosphorylation of
IB precedes its dissociation from NF-B, it is not sufficient and requires additional
ubiquitination to enable its dissociation and subsequent degradation [189].
Chapter G. Study II
54
Figure II.1 (A) A ribbon representation of the structure of a complete IB and its
phosphorylation sites bound to the partnering NF-B, composed of the RelA and p50
subunits. RelA is colored in pink and p50 colored in brown. The ankyrin repeat domain and
the C-terminal of IB are displayed in grey with the modeled signal receiving domain
(SRD) as part of the N-terminal region colored according to its secondary structure
elements with the phosphorylation sites pSer32 and pSer36 displayed in atomic detail. (B)
A multiple sequence alignment of the N-terminal regulatory region, the degron motif
constituted of residues DSGXS, with  corresponding to a hydrophobic residue (here
isoleucine) and X to any amino acid. (C) A close-up of the IB SRD displaying the modeled
helical structural elements displayed in purple. The phosphorylation sites, Ser32 and Ser36,
flank the degron motif highlighted in green.
To paint a complete picture of how phosphorylation guides and regulates protein
function, the rub lies in understanding the molecular basis of this phenomenon. With
phosphorylation being the most well-studied posttranslational modification,
computational simulation methods on the atomistic scale have proven valuable in offering
explanations of the structural basis of protein regulation by phosphorylation in addition to
complementing experimental studies [193-195].
The crystal structure of the -TrCP/-catenin complex has revealed the basis of
substrate recognition by the WD40 domain in -TrCP and shed light on the intimate
interaction between degron motif containing substrates (Figure II.1 B) and the SCF
ubiquitin ligase [56]. Of similar interest would be the interactions between -TrCP and the
phosphorylated IB; however prior to this, the subject of interest is the structural
rearrangement triggered by phosphorylation of the SRD in the ankyrin repeat protein
which drives the recognition mechanism by SCF. To this end, I have in this study examined
the role of mono- and double-phosphorylation by probing into the structural effects that it
induces on IB bound to its partner, NF-B. By applying long scale MD simulations of 500
ns, phosphorylation of Ser32 and Ser36 of IB in complex with NF-B, both in their
double- and mono-states of phosphorylation were studied. This study identified a partially
more stable region locally by the site of phosphorylation in the double-phosphorylated
IB. Moreover, double-phosphorylation induced a rather extended conformation in
Chapter G. Study II
55
vicinity of the phosphorylation site as compared to a more bent shape observed in the
unphosphorylated state. The mono-phosphorylated states each fell into an intermediary
state between the unphosphorylated and the double-phosphorylated states, only half
initiating this structural transition. In addition, distinct variations between pSer32 and
pSer36 were observed. pSer32 emerged as the more solvent exposed residue and
subsequently burying Ser36 slightly causing the formation of a new hydrogen bond pattern
stabilizing the N-terminal tail and the region closest to pSer36. The findings also revealed a
more pronounced electrostatic effect upon double-phosphorylation, which induce
structural rearrangements that change the surface charge potential and create a greater
acidic environment around the phosphorylation site. Overall, these findings explain the
prerequisite for double-phosphorylation on a detailed molecular level and offer an insight
into the structural rearrangements that take place, hence laying the ground work into
future studies of IB/-TrCP recognition and binding mode.
G.2 Methods
G.2.1 Structural model of IB/NF-B
The model of the full-length IB/NF-B complex was obtained as described in the
previous study (Study I), containing the complete signal receiving domain (SRD; residues 167) of IB and the ankyrin repeat domain (ARD; 68-280) thereby providing a suitable
working model to investigate the phosphorylation in the SRD region of IB. A model of a
double-phosphorylated simulation system is presented in Figure II.1 A.
G.2.2 Molecular simulation setup
The well-equilibrated structural model of the unphosphorylated IB/NF-B complex from
Study I of this thesis, was used as an initial structure for the simulations. To investigate the
role of phosphorylation, four distinct system setups were constructed: an
unphosphorylated (nIB) system, two mono-phosphorylated complexes where either of
serines 32 (p32IB) or 36 (p36IB) were post-translationally modified into
phosphoserines, and a double-phosphorylated system (ppIB) in which serine residues 32
and 36 were both transformed into phosphoserines simultaneously. The phosphoserine
mutations were introduced with PyMol [196]. The simulations were carried with Gromacs
4.5 [140] with the modified GROMOS96 43a1p force field [172] containing the parameters
for phosphoserine. A time step of 2 fs was used. The LINCS algorithm [173] was applied for
constraining bond lengths. Electrostatic interactions were calculated with the ParticleMesh Ewald algorithm at every step [174]. A 1.0 nm cutoff was used both for electrostatics
and van der Waals interactions, with neighborlists updated every 10 steps. The simulations
were performed at constant pressure of 1.0 bar with Parrinello-Rahman pressure coupling
[197] and the isotropic pressure scaling, time constant of 1.0 ps, and a system
compressibility of 4.5e-5 bar-1. The temperature of the system was coupled to 300 K using
the velocity-rescaling algorithm [198]. The nIB system containing 6945 protein atoms
Chapter G. Study II
56
was solvated with 31,361 SPC water molecules, 218 out of which were replaced with 122
sodium and 96 chloride ions to neutralize the total net system charge and obtain a
physiological salt concentration of 0.15 M. The total number of atoms in the wild-type
system reached 100,226 all of which were placed in a rectangular cell of an approximate
system size of 9.4 x 8.0 x 13.7 nm3. For the PTM-modified systems, sodium ions were
replaced with corresponding number of water molecules to maintain a neutralized system
charge depending on the phosphorylation state of the systems. The systems were energy
minimized with steepest descent until the maximum force reached < 100 kJ/mol/nm and
were subsequently equilibrated for 100 ps while keeping the protein position restrained (Fc
= 1000 kJ/mol/nm2). Eventually, all position restraints were removed and the production
runs were performed for 500 ns each. Three independent replicates of each system were
simulated, each starting with different initial velocities, amounting to a total simulation
time of 6 s, thus allowing a thorough exploration of the extensive conformational space.
G.3 Results
In order to study the effects of phosphorylation of Ser32 and Ser36 located on the SRD of
IB, I carried out MD simulations of the unphosphorylated and the doublephosphorylated IB in complex with the transcription factor NF-B. To further
characterize the solitary roles of mono-phosphorylation of Ser32 and Ser36, two additional
mono-phosphorylated systems were studied and compared with the unphosphorylated
and the double-phosphorylated IB/NF-B complex. The reported results are based on an
average finding of the three independent replicate runs of each system, which were
simulated for 500 ns each.
G.3.1 Local divergence in structural stability promoted by
double-phosphorylation
The degree of flexibility of the protein can be evaluated by measuring the backbone atom
root mean square fluctuation (RMSF) around the initial structure. The stretch of residues,
IB28-40, encompassing the sites of phosphorylation Ser32 and Ser36 as shown in Figure
II.2 A, reveal a sharp contrast between the unphosphorylated and the doublephosphorylated system. The RMSF of this segment varies between 2-4 Å in the
unphosphorylated system, while remaining rather stable in the double-phosphorylated
system maintaining an RMSF between 2 and 3 Å (Figure II.2 B).
Chapter G. Study II
57
Figure II.2 (A) A ribbon representation of IB, magnifying the stretch of residues IB28-40
situated on the SRD encompassing the two phosphorylation sites, SER32 and SER36. (B)
Root mean square fluctuation (RMSF) of the backbone atoms of segment IB28-40 in nIB
(black) and the ppIB (red) systems. The highlighted regions display the sites of
phosphorylation, residues 32 and 36. (C) Root mean square deviation (RMSD) of backbone
atoms for segment IB28-40 during 500 ns for the nIB and ppIB systems, calculated
relative to the initial conformation after a least-squares fitting of IB28-40. A more stable
and lower RMSD is observed in the double-phosphorylated system.
In particular around residues preceding the initial site of phosphorylation at pSer32,
residues 28 to 32, the RMSF reveals a more stable segment in the double-phosphorylated
system. The RMSF of these residues lies around 2.5-3 Å in the double-phosphorylated
system with very small fluctuations across the replicated runs as displayed by the error
bars, whereas in the unphosphorylated system a higher RMSF of 3.2-4 Å is observed with
much greater fluctuation across the different replicates. However, interestingly, the
residues preceding the second phosphorylation site, pSer36, residues 33 to 35 including
pSer36, display a somewhat lower RMSF in the double-phosphorylated system as
compared to the unphosphorylated system. The general observations in the same stretch
of residues are supported by the root mean square deviation (RMSD) of the backbone
atoms of this region (Figure II.2 C). The RMSD of this segment keeps steady at roughly 2 Å
in the double-phosphorylated complex, in comparison to the unphosphorylated system
that has an RMSD of > 2.5 Å, a difference of > 0.5 Å observed between the two systems.
Chapter G. Study II
58
In the mono-phosphorylated systems, the p36IκBα system is structurally more
stable compared to the p32IκBα system, with a sustained difference of 1 Å in the RMSF of
the residues surrounding the phosphorylation site, a trend also reflected in the RMSD in
particular in the final 300 ns of the simulations (Figure II.3). An opposing effect of the two
sites of phosphorylation is revealed in the mono-phosphorylated states when compared to
the unphosphorylated state: pSER32 modification leads to an increase in RMSF in
immediate proximity of the phosphorylation site, whereas pSER36 causes a drop in RMSF
in surrounding residues.
Figure II.3 Root mean square fluctuations (RMSF) of the backbone atoms of segment
IB28-40 (A) and the entire protein (B) in all states. (C) Root mean square deviation (RMSD)
of backbone atoms for segment IB28-40 during 500 ns for all systems, calculated relative
to the initial conformation after a least-squares fitting of IB28-40.
G.3.2 Double-phosphorylation induces an extended N-terminal
conformation of SRD
To further characterize induced structural variation due to two-fold phosphorylation, the
C-C distance between Ser32 and Ser36 was measured (Figure II.4 A). By mapping the Cα
distances between these residues, one is able to investigate in detail the effect of monophosphorylated pSer32/nSer36, nSer32/pSer36 and double-phosphorylated pSer32/pSer36
and compare with the not-PTM nSer32/nSer36 ‘unphosphorylated’ system. Chapter G. Study II
59
Figure II.4 (A) The Ser32C-Ser36C distance compared between the different simulation
systems. (B) Representative structure of the degron containing segment IB31-37 in the
unphosphorylated and (C) the double-phosphorylated system.
A remarkable deviation in the Ser32C-Ser36C distance could be observed across
the unphosphorylated and the double-phosphorylated complexes. In the doubly
phosphorylated state, the C-C distance increases gradually in the first 30 ns of the
simulation, after which it reaches a distance of 13 Å and maintains it throughout the rest of
the simulation. In contrast, the C-C distance in the not phosphorylated state drops
sharply down to 9 Å in about 30 ns and sustains this distance at about 9-10 Å for the
remaining of the 500 ns simulation duration averaged over three independent simulation
runs. In both of the mono-phosphorylated systems, the C-C distance fluctuates between
9-12 Å in the initial 300 ns of the simulation, however, into the final 200 ns both systems
keep a rather constant distance at approximately 12 Å. This shows that two-fold
phosphorylation is more effective than mono-phosphorylation of either Ser32 or Ser36
residues but that the structural effect is not additive.
This variation in the Ser32C-Ser36C distance between the unphosphorylated and
the double-phosphorylated states is exhibited in the structural conformation of the
phosphorylation region. The representative conformations of a stretch of residues 31 to 37
in the unphosphorylated (Figure II.4 B) and the double-phosphorylated (Figure II.4 C)
complexes, clearly portrait a region with a defined bend in between the sites of
phosphorylation in the unphosphorylated system, whereas a more extended structure is
revealed upon double-phosphorylation.
G.3.3 Variation in solvent exposure in Ser32 and Ser36
The impact of phosphorylation on sites Ser32 and Ser36 can be assessed by calculating the
relative solvent accessible surface area (SASA) of the individual serine residues.
Chapter G. Study II
60
Figure II.5 Relative solvent accessible surface areas (SASA) of Ser32 and Ser36 in the
unphosphorylated (A), the pSer36 mono-phosphorylated systems (B). pSer32 monophosphorylated (C), and double-phosphorylated (D); Ser32 is depicted in cyan, and Ser36 is
colored in purple.
As evident from Figure II.5 A, in the unphosphorylated complex, both Ser32 and
Ser36 remain surface exposed (staying above a threshold of 20 %), with Ser36 holding the
upper hand with a slight margin fluctuating around a relative SASA of 40 %. Intriguing are
also the effects of mono-phosphorylation on each respective unphosphorylated partner.
Mono-phosphorylation of pSer32 sets in motion structural rearrangements, which in turn
induce a partial burial of Ser36 as shown in Figure II.5 B. However, the monophosphorylation of pSer36 triggers firstly, the increase of the relative SASA of pSer32 as
compared to the unphosphorylated structure, and secondly, it allows a higher surface
exposure of pSer36, even higher than the pSer36 relative SASA in the doublephosphorylated state (Figure II.5 C). In clear contrast to the unphosphorylated state,
double-phosphorylation leads to a striking difference in the relative SASA between pSer32
and pSer36 (Figure II.5 D). Although phosphorylated, pSer36 maintains a similar but less
fluctuating relative SASA of 40 %, suggesting partial burial of this residue. However, pSer32
maintains a highly exposed configuration with a relative SASA of 80 % during a larger part
of the simulation, allowing the potential for a wide range of interactions. These
observations are suggestive of the distinctive roles that pSer32 and pSer36 could play upon
phosphorylation: a rather exposed pSer32 which extends beyond the surface of the protein
to act as an anchoring point and engage in inter-protein interactions, and a comparatively
less exposed pSer36 that could be contribute to key-specific intra-protein interactions.
Chapter G. Study II
61
G.3.4 Double-phosphorylation stabilizes region by novel
hydrogen bond interactions
The conformational differences caused by phosphorylation may be accounted for by a new
hydrogen bond formation pattern in the double-phosphorylated complex. Based on our
simulations, out of the two phosphorylation sites, it is in particular pSer36 that contributes
to establishing newly formed interactions.
Figure II.6 (A) Cartoon representation of double-phosphorylated IB1-70 color coordinated
according to secondary structure elements. The green shaded segment flanked with the
red colored spherical phosphoserines denotes the degron in IB. The N-terminal, depicted
in yellow, is shown to interact closely with pSER36 that forms hydrogen bonds with
residues Met1 and Gln3 situated right by the N-terminal tail. Additional unique hydrogen
bonds formed upon phosphorylation in the vicinity of the phosphorylation site involve
residues Asp35 and Gln44 (B) and Met37 and Glu43 (C). The hydrogen bond interactions
deemed significant are present in at least 30 % of the simulation period in at least 2 out of 3
of the replicate runs.
Residues Met1 and Gln3 located at the tip of the N-terminal shift away from pSer32
and closer to pSer36 and form unique hydrogen bonds: The backbone amide of Met1
interacts with the phosphate group of pSer36, whereas both the backbone and side chain
amides of Gln3 form hydrogen bonds with the backbone carbonyl group of pSer36; these
shifts together with other local conformational changes lead to the stabilization of the Nterminal tail and concomitant other regional changes in the structure of IB (Figure II.6 A).
Additional residues that come to form hydrogen bonds in vicinity of the phosphorylation
site are Asp35/Gln44 (Figure II.6 B) and Met37/Glu43 (Figure II.6 C). The backbone carbonyl
of Aso35 engages in hydrogen bond interaction with the side chain amide of Gln44, and the
Chapter G. Study II
62
backbone amide of Met37 forms a similar interaction with the side chain carboxyl group of
Glu43. Interestingly, Asp35 and Met37 both enclose the pSer36 phosphorylation site, and
the adjoining residues Gln44 and Glu43 are localized on the third -helix in the SRD of
IB. This newly formed interaction pattern present uniquely in the doublephosphorylated while absent in the unphosphorylated states, highlights the role that
phosphorylation possesses in establishing stability and order in the region, as reflected
previously in the lower RMSF exhibited by the double-phosphorylated state. Specifically, it
appears that phosphorylation of Ser36 is the principal contributor to this effect.
G.3.5 Electrostatic effects
Part of the effects of phosphorylation may be explained by differences in electrostatic
potential surrounding the site of phosphorylation. The degron motif in IB31-36 holds the
sequence of residues DSGLDS, which encompasses the integral site of phosphorylation
Ser32 and Ser36 each of which are preceded by an acidic residue, aspartic acid. The
electrostatic potential of this region is illustrated in Figure II.7, with red colored surface
correlated with potentials of -10 kT and blue colored surface correlating with potentials of
+10 kT.
Figure II.7 Electrostatic potentials of the SRD mapped onto the van-der-Waals protein
surface of IB calculated by APBS [199] in the nIB (A), pSer36IB (B), pSer32IB (C)
and the ppIB states (D). The position of the site of phosphorylation is indicated by the
cartoon figure of segment IB31-37. (E) The electrostatic surface potential on the WD40
domain of -TrCP, showing a top view of the binding interface to a double-phosphorylated
IB. Negative potentials of -10 kT are depicted in red, and positive potentials of +10 kT are
depicted in blue. The residues making intermolecular contacts with the phosphoserines in
the degron motif are indicated in their respective positions.
In the unphosphorylated state (Figure II.7 A), the negative potential patches at or
near the protein surface surrounding the sites of phosphorylation are fostered by the
presence of strong acidic residues, namely Asp27, Asp28, Asp31 in the -helical region
Chapter G. Study II
63
preceding Ser32 and residues Asp39, Glu40, Glu41, Glu43 that are localized on the loop
segment following Ser36. In the double-phosphorylated state, the previously negatively
charged patch grows even stronger covering a more extended part of the protein surface
(Figure II.7 D). The introduced negative charges upon phosphorylation, jointly with induced
structural rearrangements alter the distribution of surface charges potential in a visible
manner creating an even more pronounced negatively charged protein surface. The monophosphorylated states, as displayed by Figures II.7 B and C, do not exhibit the same
electrostatic effects as their double-phosphorylated counterpart. Since, doublephosphorylation is a prerequisite for -TrCP binding, the electrostatic surface potential of
the binding area of -TrCP is shown (Figure II.7 E). The surface potential of the top narrow
part of the channel being the binding surface reveals an extensive positive blue colored
patch indicative of a basic environment. The electrostatic complementarity of doublephosphorylated IB and -TrCP protein surfaces may be the critical recognition
mechanism to initiate the formation of the tertiary protein complex.
G.4 Discussion
Phosphorylation is the most prevalent post-translational mechanism regulating protein
function throughout the cell. Protein kinases carry out phosphorylation by modifying
existing serine, threonine and tyrosine side chains with the addition of a phosphate group.
At physiologically relevant pH, a phosphate group usually carries a -2 charge. Introducing
such a negative charge often leads to electrostatic perturbations directly affecting protein
energy landscapes, which exercise control over protein-protein interactions and
conformational dynamics [200]. As with many other cellular functions, it is challenging to
derive an atomic-level understanding of how phosphorylation alters protein structure and
function. Computational methods, such as all atomistic MD simulations, are a major
contributor in filling this knowledge-gap and throwing light on the structural changes upon
protein phosphorylation (as reviewed in [200]).
Many experimental studies on the importance of IB phosphorylation have been
done in the past [47,188-192]. Yet little is known about the molecular details, which render
this post-translational modification indispensible for recognition by the SCF complex. With
MD simulations, I was able to characterize local structural aspects induced by
phosphorylation of Ser32 and Ser36. Although no striking differences were observed in
secondary structure elements upon phosphorylation, aspects of local stability, in the
progression of disorder to order surrounding the sites of phosphorylation were evidenced.
Double-phosphorylation leading to the uniform ordering of the segment IB28-40 was in
part brought forth by the stabilization of residues preceding the initial phosphorylation site,
namely residues 28-32. Moreover, increased N-terminal stabilization induced by forming
hydrogen bond interactions with the phosphorylation site was another direct effect of
double-phosphorylation. In contrast, the unphosphorylated IB experienced a more
variable structural stability across this region. Post-translational phosphorylation has often
previously been associated with sparking both ordering and disordering in proteins. For the
degron of IκBα in complex with NF-kappaB, we observe an opposite effect which was also
observed, for example for smooth muscle myosin [201]. Such a residue-to-residue decrease
in phosphorylated serine amino acid residues was also observed in the α-amino-3-hydroxy-
Chapter G. Study II
64
5-methyl-4-isoxazolepropionic acid (AMPA) receptor and interpreted as a coil-to-helix
compressed secondary structure change [202].
A Monte Carlo/Stochastic Dynamics simulations study revealed how electrostatic
interactions navigated the stabilization of a helical conformation at the N-terminus upon
phosphorylation in model peptides consisting of a serine in the N-terminus followed by
nine alanines [203]. Via docking methods, phosphorylation of the phenylalanine
hydroxylase (hPAH) demonstrated in another study an increase in stability of the Nterminal tail through local conformational changes as a result of electrostatic interactions
[204]. In an investigation employing the Car-Parrinello ab initio MD method,
phosphorylation of the CREB-CBP complex involved in activation of the DNA transcription
machinery results in stability of the complex by a new hydrogen bond interaction [205]. In
another study employing molecular mechanics–based methods on kinases and prokaryotic
response regulators, local conformational changes specific to areas in close proximity to
the phosphorylated amino acid were observed [206]. Structural transitions between order
and disorder induced by phosphorylation have also been mapped by previous experimental
work. An order to disorder transition of a helix containing a phosphorylated serine was
documented in the Oncoprotein 18/stathmin (Op18) phosphoprotein [207], in contrast to a
disorder to order adaptation of the protein kinase B/Akt which, after phosphorylation,
experiences disruption of the C-helix causing global restructuring of the protein [208].
The unphosphorylated state simulations in this study indicated a bent shape
between the two serine residues. However, an impact of double-phosphorylation was the
structural reconfiguration of the segment connecting the phosphoserines resulting in an
extended conformation and verified by a larger Ser32C-Ser36C distance. Findings in a
study by Pons et al. [209] proposed a random coil conformation withholding a weaker bend
of a free phosphorylated 24 amino acid segment, IB21-44 characterized by NMR
spectroscopy and MD simulation, in contrast to the conformation of the bound segment to
-TrCP obtained by transfer nuclear Overhauser effect spectroscopy (tr-NOESY)
presenting an apparent bend between the two serines. The occurrence of a well-defined
bend connecting pSer32 and pSer36 in the bound structure could be initiated by structural
rearrangement upon binding to the ubiquitin ligase.
A distinct pattern of conduct of the phosphorylated serines, revealed by the
simulations, is indicative of the individual roles played by each serine residue. I propose that
phosphorylation is commenced at site Ser36, which in turn increases the solvent surface
exposure of Ser32 increasing its accessibility to the kinase IKK, although partially burying
pSer36 (Scheme II.1).
Scheme II.1 A hypothetical model of the sequential phosphorylation of IB by the protein
kinase IKK. Based on these simulations, Ser36 is phosphorylated first by the IKK, which in
Chapter G. Study II
65
leads to an increased solvent accessible surface area of the second phosphorylation site,
Ser32. With an enhanced exposure to the solvent, Ser32 is in turn phosphorylated by the
kinase, with the effect of an extended structural loop conformation of the degron segment.
This relative partial burial of pSer36 leads to a new hydrogen bond network
stabilizing the N-terminal segment of IB and the region proximal to the pSer36
phosphorylation site. With pSer36 contributing to local structural stability of the protein,
pSer32 with a relatively large SASA is able to interact freely with -TrCP. This view is
supported by the resolved crystal structure of the bound -TrCP to -catenin showing how
pSer33 (homologous to pSer32 in IB) is the residue that makes the largest number of
contacts with -TrCP, that is residues Tyr271, Ser309, Ser325 and Arg285, while pSer37
(homologous to pSer36 in IB) forms comparatively fewer interactions with residues
Ser448, Gly432 and Arg431 (Figure I.7 E) [56]. These residues are located on the rim at
opposite sides of the channel. Moreover, molecular docking data of a phosphorylated 11
amino acid IB peptide bound to -TrCP demonstrated that pSer32 establishes the same
interactions as -catenin [210].
Long-range electrostatic effects are a major determinant of protein-protein
recognition and association. The electrostatic complementarity of protein surfaces was
identified to be a major regulator of protein-protein complex formation, see for example
[211-213]. Post-translational introduction of phosphate groups carrying a double negative
charge by IKK enhances the negative electrostatic potential in the double-phosphorylated
state of IκBα, an effect remarkably different from the unphosphorylated and the monophosphorylated states. The elicited negatively charged protein environment adjacent to
the phosphorylation sites is also of critical importance in the recognition by the -TrCP SCF
complex.
Chapter H. Study III
66
Chapter H
Chapter H. Study III
67
Study III
H.1 The Molecular Basis of Polyunsaturated Fatty Acid
Interactions with the Shaker Voltage-Gated Potassium
Channel
Polyunsaturated fatty acids (PUFAs) being essential parts of cell membrane phospholipids
of heart cells and neurons [214] have been found to shift the voltage dependence of the
Shaker voltage-gated K+ (Kv) channel to enable channel activation via a proposed
electrostatic mechanism [108,128]. While the fatty acid modulatory effects have been
extensively studied, in particular on voltage-gated ion channels [215], the molecular
mechanism by which PUFAs modulate channel functioning remains an open question.
PUFA molecules are embedded within the lipid bilayer, and the modulation of Kv
channel activity can in principle occur via indirect or direct effects, or a combination of the
two. For example, addition of PUFAs to cardiac myocytes blocked sodium currents in Na+
channels by increasing the membrane fluidity, which is suggestive of an indirect effect
[104]. On the other hand, direct PUFA interactions have been observed for several voltagegated ion channels [103,105,107,125-127] and are often inferred by exclusion of indirect lipid
bilayer effects. In this way, direct PUFA-mediated inhibition have been observed in Kv1.5
channels [103], NaV channels [125,127], and CaV channels [126] and both activation and
inactivation on the Kv1.5 and Kv2.1 channels [105]. In addition, mutational analyses
pinpointed direct effects of PUFAs to inhibit NaV channels targeting the S6 helix in the pore
[107] and to activate Kv channels via the extracellular part of the VSD [129].
Chapter H. Study III
68
Because Kv channels are integrated components of the nervous system, their
malfunction is often connected to disease. As a result, refractory epilepsy [110-112], which
is triggered by Kv1-type or Kv7-type channel malfunction [119-124], can effectively be
treated by ketogenic diets containing a high PUFA content [108,110]. Even though
ketogenic diets have been prescribed to patients since the 1920s [216], the underlying
mechanisms by which it operates and prevents the epileptic seizures remain elusive.
However, given the wide scope of PUFA effects on a range of voltage-gated ion channels,
the mechanism of ketogenic diets likely include PUFA-channel interactions.
The PUFA head group (Figure III.1 A) was observed to activate Shaker Kv channels,
which is referred to as the lipoelectric mechanism [128,217] and can be abolished or
reversed by neutralization of the negative PUFA head group charge or introducing a
positive charge [128]. Similarly, the acyl chain properties have also proven important. While
the acyl tail length does not seem to be critical, the number and geometry of the double
bonds have significant effect on the modulatory properties of the PUFA [128], e.g. a
minimum of two double bonds, particularly in the cis arrangement, substantially increase
channel currents. PUFAs have been reported to modulate the gating mechanism of the
Shaker Kv channel by partitioning into the lipid bilayer and interacting with the extracellular
halves of the VSD helices S3 and S4 [129]. Specifically, a series of cysteine point mutations
on the S3, S4, S5, and S6 helices revealed that residues on helices S3-S4 altered the
sensitivity to the docosahexaenoic acid (DHA) PUFA. In addition, introducing positively
charged cysteine-specific MTSEA+ probes identified four high-impact residues on the lipidfacing side of the VSD cytoplasmic region; I325 and T329 (helix S3) and A359 and I360 (helix
S4) (Figure III.1 B). Indeed, channel VSDs make significant contacts with the surrounding
lipid bilayer [97,98,218] and MD simulations show specific interactions between the gatingcharge R1 and R2 arginines on helix S4 on Kv1.2 channels to salt-bridge to lipid headgroups
[85]. In addition, S4 gating charges have been observed to interact with lipid head groups in
open, resting and intermediate states of the Kv1.2 and the paddle–Kv1.2 chimera channels
[219-221]. Finally, the S3b-S4 paddle in VSDs of Kv channels has been identified as the key
interaction point between the lipids and the channel [222]. Hence, a picture emerges where
charged PUFAs partitioned into the lipid bilayer exert modulatory effects by contributing
specific interactions with the VSD domains of ion channels.
Chapter H. Study III
69
Figure III.1 Molecular PUFA-VSD interactions in the Shaker channel. (A) Close-up of a PUFA
(DHA) in its initial conformation. The numbers marked in grey depict the carbons forming
the cis double-bonds. (B) Side-view of one VSD equilibrated in a POPC lipid bilayer
represented by a yellow iso-density surface corresponding to the positions of lipid
nitrogens in the simulation at 5 % occupancy. The residues shown in experimental studies
to be close to the interaction site of DHA, namely residues I325, T329 located on S3, and
A359, and I360 located on S4, are colored in cyan [129]. (C) Top-view of the Shaker tetramer
with PUFAs in their starting positions. The PUFA carboxyl head group and carbon tail are
colored in blue and green, respectively. The simulated dynamics of the PUFAs surrounding
the Shaker tetramer is represented by a brown mesh iso-density surface at 27 % occupancy.
The cut-off was chosen to visualize the differences between the PD and VSD interactions
with the PUFAs.
This study characterizes PUFA-Kv channel interactions in the open and closed states
of the channel using atomistic MD simulations. The study explores the interaction between
the Shaker Kv channel in an open state and a PUFA-enriched lipid bilayer and specifically
characterized PUFA enrichment regions on the VSD. The open state Shaker Kv channel was
modeled based on the high-resolution experimental structure of the chimera channel
Kv2.1/Kv1.2 [82]. A potential PUFA-Kv channel site of interaction was found to be located on
the lipid-facing side of a pocket connecting the extracellular halves of S3 and S4 helices,
which is supported from experiments. In general, the lipophilic PUFA tail covered a wide
range of non-specific hydrophobic interactions along helices S3 and S4, while the carboxylic
head group formed fewer and more specific electrostatic interactions with the top regions
of the S3-S4 helices and the S3-S4 linker. In addition, by performing simulations of
saturated fatty acids (SFA)-Kv channel systems, the prerequisite of a polyunsaturated
carbon tail is explained by simulations as suppression of flexibility in a saturated carbon tail.
The closed state simulations revealed an interaction pattern in which both the PUFAs and
SFAs formed fewer interactions compared to the corresponding open state simulations.
Together, the results explain the selective stabilization of the open state of a Kv channel,
identify a putative PUFA interaction site at atomic detail and thereby provide novel K v
Chapter H. Study III
70
channel interaction points that can be tested experimentally and aid in the design of
pharmaceutical compounds for the treatment of epilepsy.
H.2 Methods
H.2.1 Shaker channel system setup in open state
The Kv Shaker channel has been extensively studied with respect to PUFA modulation
(gating function) [108,128,129]. Because of a 66 % sequence identity in transmembrane
(TM) helices S0-S6 between the Shaker K+ channel and the high-resolution X-ray structure
of the Kv2.1 paddle–Kv1.2 chimera channel (PDB ID 2R9R) [82], the structural model of the
Shaker in an open state was based on this structure. A Shaker channel with partially
truncated S3-S4 linker (residues 337-350) has been shown to have similar biophysical
properties to the wild-type channel [223,224]. Therefore, I opted for removing this
extended loop, which is also missing in the chimera crystal structure. The threedimensional structure of the Shaker channel was constructed with Modeller 9.12 [136].
Initially, five models were built and the optimal model was identified using a combination
of the following model quality scores: The Discrete Optimized Protein Energy (DOPE)
score, a statistical potential used to assess homology models [225], and the GA341 score, a
method used for model assessment based on the percentage sequence identity between
the template and the model [226].
The ion channel was immersed in a pure POPC bilayer [227] consisting of 256 lipids
per leaflet using the g_membed tool in the GROMACS tool repository [228]. After inserting
the protein into the bilayer 438 POPC lipids remained. The system was energy minimized
with steepest descent until the maximum force reached < 1000 kJ/mol/nm and was
subsequently equilibrated for 10 ns while keeping the protein and the waters position
restrained (Fc = 1000 kJ/mol/nm2). With the position restraints removed, the system was
relaxed in a 50 ns simulation. The Amber ff99SB-ILDN force field [229] was used for the
construction of the membrane-protein system, in combination with the Berger force field
[227] for the lipids and the TIP3P water model [230] and the CHARMM36 force field [231]
was used for the 1 μs production run.
H.2.2. PUFA and SFA system setup in open state simulations
Force field parameters for the PUFA (docosahexaenoic acid), SFAs (docosanoic acid) and
the MTSEA+ compounds were obtained from the multipurpose atom-typer for CHARMM
(MATCH) server [232], which utilizes libraries of topology and parameter files in existing
force fields for extrapolation to the new molecules consistent with the parameterization
strategy within a given force field. The default Charmm General Forcefield
(top_all36_cgenff) was used for deriving partial charges and parameterization. The
calculated order parameters for DHA (Figure III.2 A) are in agreement with those obtained
previously both by computational and experimental efforts [233]. In addition, the order
parameters for the SFA agree with the increased level of disorder towards the end of the
Chapter H. Study III
71
chain reported for the saturated palmitic acid [233]. Together, these comparisons validate
the parameterization procedure.
The simulations were performed using the all-atom CHARMM36 force field [231]
and a development version of Gromacs [170,171]. The LINCS algorithm [173] was applied
for constraining bond lengths. Electrostatic interactions were calculated with the ParticleMesh Ewald algorithm at every step [174]. P-LINCS [234], a non-iterative parallel
constraints algorithm allowing replacement of hydrogens with virtual interaction sites,
enabled 5 fs time steps [140]. A 1.2 nm cutoff was used both for electrostatics and van der
Waals interactions, with neighborlists updated every 10 steps. The simulations were
performed at constant pressure of 1.0 bar with Parrinello-Rahman pressure coupling [197]
and the semiisotropic pressure scaling, time constant of 2.0 ps, and a system
compressibility of 4.5e-5 bar-1. The temperature of the system was maintained at 300 K
using the velocity-rescaling algorithm [198].
PUFAs were placed around each VSD on each lipid leaflet (Figure III.1 C) and
clashing POPC lipids were removed resulting in 80 PUFAs and 275 POPC lipids surrounding
the tetrameric Shaker channel. The system was subsequently solvated with 33,389 TIP3P
waters and 192 sodium and 104 chloride ions replaced water molecules to neutralize the
net system charge (and obtain a salt concentration of 0.1 M). The total number of atoms in
the system reached 163,139. Initially, the system was equilibrated for 30 ns while keeping
the protein, waters, and PUFAs frozen. Position restraints were then applied to the protein,
PUFAs, and z-coordinates of the waters followed by 100 ns equilibration. Hereafter, all
position restraints were removed except for the PUFAs and the system was allowed to
equilibrate for an additional 10 ns. This series of equilibration steps allowed POPC lipids to
pack around the PUFAs while preventing water molecules from penetrating into the
membrane bilayer. The SFA-channel system was created by replacing the PUFAs in this
equilibrated system with SFAs followed by a 200 ps equilibration step.
Two initial PUFA distribution patterns were tested. In the first, 32 PUFAs were
evenly distributed across the lipid bilayer (Figure III.3 A) and sampling was performed in a 5
μs production run. In the second setup, center of mass (COM) pulling was applied using an
umbrella potential between each VSD and the surrounding PUFAs in order to allow for
close (but non-specific) protein-PUFA packing. Here, the PUFAs were pulled in the x and y
dimensions with a harmonic force constant of 1000 kJ mol-1 nm-2 at a rate of 0.08-0.01
nm/ps depending on the initial distances between the PUFAs and the VSD followed by a 1
μs production run. The COM pulling procedure was repeated to create the SFA system.
H.2.3 Modeling of Shaker channel in closed state
A model of the Shaker channel in a closed state was built based on a previous Rosetta
model of the channel in the C3 state [235] with a partially truncated S3-S4 linker using
Modeller 9.12 [136]. This closed-state model replaced the open state channel in the
previously set up PUFA and SFA systems. The new system configurations were relaxed for
10 ns followed by COM pulling to position PUFA and SFA molecules in initial positions close
to the channel and subsequent 1 μs production simulation runs. To assess the structural integrity of the structures, the root mean square deviation
(RMSD) of the protein backbone was calculated against the average structure for all four
systems (PUFA and SFA channel systems in open and closed states) and compared to a
Chapter H. Study III
72
channel-only system. Overall, the channels maintain stable structures, settling around an
RMSD of 1.5 – 2.0 Å throughout the simulation time.
H.2.4 In silico mutagenesis of the VSD
Residues I325, T329, A359, I360, and L366 were mutated to cysteines and modified to
include the reagent MTSEA+ to characterize PUFA interaction sites. The mutations were
carried out using PyMOL (The PyMOL Molecular Graphics System, Version 1.3
Schrödinger, LLC). Each mutated system was minimized with a steepest descent
algorithm until a maximum force of < 1000.0 kJ/mol was reached followed by a short 200 ps
equilibration. Production runs were 1 μs long for each of the MTSEA+ mutated systems,
except for the L366 system which was simulated for 500ns. In addition, the MTSEA +
modified systems were mutated back to cysteines after 500 ns of simulations and
continued for 500 ns.
H.2.5 Small-scale PUFA and SFA systems
Two membrane patches were built using the membrane generator MemGen [236] with 24
lipids per leaflet. In each of the bilayer patches, DHA or DA, were inserted in each lipid
leaflet and were set up as the described PUFA system. The systems each contained ~12,000
atoms with roughly 48 POPC lipids, 1,660 water molecules and 20 sodium and chloride
counter ions. Each system was subjected to steepest descent energy minimization until the
maximum force reached a value < 1000 kJ/mol/nm followed by 1.5 ns of equilibration.
Production runs were 1 μs for each system configuration.
H.3 Results
In a set of independent simulations, I explored dynamics of polyunsaturated
(docosahexaenoic acid, DHA) (Figure III.1 A) and saturated (docosanoic acid, DA) fatty acids
in neat lipid bilayers as well as bilayers containing the Shaker Kv channel in both its open
and closed states.
H.3.1 Saturation levels in the carbon tail affect the structural
dynamics
To determine how saturation levels in the fatty acid tail influenced the general structural
dynamics, single PUFA (DHA) and SFA (DA) molecules in two separate 1-palmitoyl-2oleoyl-sn-glycero-3-phosphocholine (POPC) membrane patches were simulated. As
expected, the observed order parameters for both the PUFA and SFA molecules decreased
gradually from the carboxyl end located in the bilayer interface to the end of the acyl chain
tail in the bilayer center (Figure III.2 A). While the order parameters were quite similar near
the carboxyl head and at the methyl end of both fatty acids, the PUFA order parameters
were generally lower and displayed a different overall shape reflecting the positions of the
Chapter H. Study III
73
cis double-bonds. Thus, the PUFA molecule exhibited greater conformational mobility
compared to SFA in the POPC membrane patch.
Figure III.2 Structural flexibility in lipid-partitioned PUFA and SFA. (A) Deuterium order
parameters of PUFA (green) and SFA (purple) carbon chains. (B) The radius of gyration of
PUFAs (green) and SFAs (purple). (C) Distances between the head group oxygen and the
last carbon in the PUFA (green) and SFA (purple) chains. The dotted lines denote the
overall average distance for the PUFAs (black) and the SFAs (blue), respectively.
Representative structures from 1 µs simulations of PUFAs (D) and SFAs (E) are shown
embedded in POPC bilayers.
Furthermore, I characterized differences in the overall shape and packing properties
by measuring the radius of gyration of the carbon tail and head-to-tail length. The
polyunsaturated chains displayed a significantly lower radius of gyration as compared to
the fully saturated acyl chains (Figure III.2 B). Similarly, the distance between the carboxyl
head group oxygen and the final methyl carbon in the tail (O-C22 distance) was 5 Å shorter
in PUFAs compared to the SFAs (Figure III.2 C). In addition, the fluctuations in the O-C22
distance were more pronounced in the polyunsaturated carbon chains with extreme values
Chapter H. Study III
74
between 10 Å and 20 Å. Hence, while the PUFA molecule tends to twirl and curl up (Figure
III.2 D), the average SFA conformation is more extended (Figure III.2 E).
H.3.2 PUFA interactions with the Kv Shaker channel in the
open state
To characterize PUFA-protein interactions in the membrane, I initially distributed 32 PUFAs
evenly across a POPC bilayer containing the Shaker tetramer, 16 on each bilayer leaflet
(Figure III.3 A) and simulated for 5 μs (Figure III.3 B). Amino acids residues that resided
within 3.5 Å of the PUFAs for more than 300 ns were mapped out in order to visualize the
region of interaction between the channel and PUFAs on the outer leaflet of the
membrane. The observed interactions differed significantly between the PUFA head and
tail groups with the PUFA tails exhibiting approximately twice the number of contacts
compared to the head groups. The contacting residues were distributed on the S1-S2 and
S3-S4 segments of the VSD. Residues close to the PUFA head groups were either charged
or polar and located on the outer regions of the VSD whereas residues contacting tail
groups were predominantly hydrophobic and covered the inner parts of the S2 and S3
helices (Figure III.3 C). While this approach provided us with an initial idea of the PUFA
interaction patterns, translational diffusion might prevent identification of realistic PUFA–
channel interactions.
Chapter H. Study III
75
Figure III.3 PUFA- Shaker interactions in a 5 μs MD simulation. (A) Top-view of the Shaker
tetramer with 16 PUFAs on each leaflet of the membrane. The PUFA carboxyl head group
and carbon tail are colored in blue and green, respectively. (B) Lateral displacement (x,y
dimensions) of PUFAs in the membrane bilayer over 5 μs with the protein COM centered in the box. (C) Interacting VSD residues presented separately for the PUFA carboxyl head
groups (blue) and carbon tails (brown).
To increase sampling, I opted for a new system configuration starting with PUFAs
packed around the VSD of each monomer in a semi-circular fashion (Figure III.1 C) using
steered MD simulations. Starting from these initially closely packed positions, extensive
sampling would in principle allow differentiation between specific and non-specific binding
by monitoring PUFA structural dynamics given a PUFA diffusion constant of 4 x 10-9 cm2/s
observed in the simulation. The resulting simulation trajectory showed the PUFAs to form
clusters in the vicinity of helices S3 and S4 across all VSD subunits rather than pore helices
S5-S6 (Figure III.1 C, green mesh surface). To identify interaction patterns, amino acid
residues that resided within 3.5 Å of the PUFAs in the outer lipid leaflet for more than 300
ns were compared between the PUFA carboxyl head groups and carbon tails (Figure III.4).
All the reported contacting residues were positioned on the VSD and not on the pore
domain. In addition, the interactions were significantly different between the PUFA head
group and tail. While the PUFA carboxyl head groups engaged in fewer contacts but with
Chapter H. Study III
76
higher contact frequencies (Figure III.4 A), the tails displayed a wider range of contacts
occurring with lower frequencies (Figure III.4 B).
Figure III.4 PUFA contacts to Shaker tetramer residues in the open and closed states. The
contact frequencies of amino acid residues within 3.5 Å of PUFA carboxyl head groups and
carbon tails are displayed for the open (A and B) and closed (C and D) states. The red dotted
line differentiates contacts on helices S1 and S2 or helices S3 and S4 of the VSD. Side-view
of a VSD and interacting residues are displayed separately for the PUFA carboxyl head
groups and tails for each state of the channel (insets).
Protein residues contacting the PUFA tails were predominantly hydrophobic and
distributed across the extracellular halves of the S1-S2 loop and helices S3 and S4 with 85 %
of the contacting residues localized on the extracellular halves of S3 and S4 (Figure III.4 B).
The PUFA tail interactions to the protein displayed lower contact frequencies compared to
the head groups, which indicates non-specific lipophilic interactions. In contrast, the
majority of the contacts between the PUFA head groups and the protein were located on
the top parts of the extracellular halves of S3 and S4, incorporating the S3-S4 linker and
were either charged or polar residues displaying high contacting frequencies (Figure III.4 A).
Specifically, the S3-S4 linker residues S351 and N353 formed hydrogen bonds with the
PUFA head groups. In addition, the negatively charged carboxyl groups of residues E333,
E334, and D336 located on the S3-S4 linker interacted with the negatively charged PUFA
head groups mediated by Na+ ions. Presence of 100 mM NaCl neutralized the simulation
system and sodium ions are typically observed in the vicinity of the protein. An additional
PUFA-protein electrostatic interaction with relatively high contact frequency is mediated
by R362, which is one of the gating charges (R1). Parenthetically, I also observed
Chapter H. Study III
77
interactions between the S4 gating charges R365 (R2), and R368 (R3) and one PUFA head
group. These interactions were enabled by a nosediving movement of that particular PUFA
molecule into the water-filled extracellular crevice. Because this PUFA-protein interaction
appeared only once in our simulations, I will not ascribe significant physiological relevance
to the interaction per se, but rather point out that PUFA interactions to the gating charges
are structurally possible in a dynamic system.
H.3.3 PUFA interactions with the Kv Shaker channel in the
closed state
To investigate the PUFA interactions with the Kv Shaker channel in the closed state. I
generated a system consisting of a closed-state model with PUFA molecules in similar
starting configurations as for the open-state system and simulated for 1 μs. The observed packing environment differed in-between PUFAs in the open and closed states of the
channel as displayed by the results of contact analyses. In general, significantly fewer PUFA
interactions were made in the closed-state simulation with an equal distribution of the
number of contacts between the head and tail groups (Figure III.4 C and D). In addition,
only half of these contacts were observed in the open-state simulation. To further
differentiate the PUFA interaction pattern between the open and closed states of the
channel, I measured minimum distances between different segments of the fatty acids and
the channel (Figure III.5). The major differences in minimum distances to the channel were
found in the head groups with PUFA carboxyl head groups being 0.5±0.2 Å closer in the
open state. In contrast, no significant variations between open and closed states were
observed in the minimum distances between the channel and different sections of the
PUFA tails.
Figure III.5 The average minimum distances between the channel in its open and closed
states and the PUFAs and SFAs. The fatty acids were sectioned into five parts consisting of
carbons C2-6 (TAIL-A), C7-11 (TAIL-B), C12-16 (TAIL-C), C17-22 (TAIL-D), and the head group
(HEAD). The error bars indicate the standard error of the minimum distances across the
four subunits of the channel.
Chapter H. Study III
78
H.3.4 SFA interactions with the Kv Shaker channel in the
open and closed state
Because SFAs do not exert the modulatory effects of PUFAs, I explored SFA interactions
with the channel in the open and closed states. The minimum SFA-channel distances
showed a pattern that was identical to that observed for the PUFAs (Figure III.5). The SFA
carboxyl head groups were on average 0.4±0.1 Å closer to the channel in the open state,
while no significant differences were detected across the different sections of the tail.
However, the numbers and contact frequencies of interactions made by the SFA carboxyl
head groups and carbon tails to the protein differed significantly between the open and
closed states. SFA interactions for both head and tail regions in the closed state were lower
both in number and contact frequency compared to the open state interactions (Figure
III.6).
Figure III.6 SFA contacts to the Shaker tetramer in the open and closed states. The contact
frequencies of amino acid residues within 3.5 Å of SFA carboxyl head groups and carbon
tails are displayed for the open (A and B) and closed (C and D) states of the channel. The
red dotted line differentiates between helices S1 and S2 or helices S3 and S4 of the VSD.
Side-view of a VSD and interacting residues are displayed separately for the PUFA carboxyl
head groups and tails for each state of the channel (insets).
Chapter H. Study III
79
Interestingly, the specific protein residues involved in PUFA and SFA interactions
differed significantly. While almost all protein residues contacting PUFA head groups in the
open state were located on helices S3-S4, the contacting residues were shifted to helices
S1-S2 for the SFAs (Figure III.4 A and III.6 A) and the PUFA tail interactions in the open state
were more numerous than for SFAs (Figure III.4 B and III.6 B). In addition, about half of the
contacts between the protein and the SFA head and tail groups were completely unique.
Finally, the specific contact patterns seem independent of the overall dynamics since there
were no significant differences in residency times between PUFAs and SFAs in neither open
nor closed states (Figure III.7). However, while the distribution of the POPC residency times
indicated more dynamics compared to PUFAs, this difference was reduced in the vicinity of
four VSD residues shown experimentally to be involved in PUFA modulation of the Shaker
channel [129]. In this position, three POPC molecules accompanied one single PUFA and
reported 1 μs residency times.
Figure III.7 Number of PUFA/SFA/POPC molecules within 2 Å of the channel. Residence
time for the number of PUFA, SFA (A) and POPC molecules (B) in the open and closed
Chapter H. Study III
80
states of the channel. (C) Residence time for the number of PUFA and POPC molecules in
the open state of the channel in vicinity of four VSD residues shown experimentally to be
involved in mediating the PUFA effect.
H.3.5 Electrostatic interactions characterize the PUFA
interaction sites
In the open state PUFA simulation, PUFAs were observed to cluster in the lipophilic pocket
of the VSD at the extracellular end of the channel proximal to the S3-S4 linker with the
hydrophobic tail tucked between helices S3 and S4 (Figure III.8 A). To characterize the
identified PUFA interaction sites, I turned to reproduction of experimental data. Mutating
residues I325 (S3), T329 (S3), A359 (S4), and I360 (S4) to cysteine and subsequent
modification using positively charged MTSEA+ affect the PUFA-induced shift of the Shaker
Kv channel [129]. Therefore, these four residues were proposed to directly modulate
function by interactions to PUFA molecules. I introduced positively charged MTSEA+
cysteine labels at positions I325, T329, A359, and I360 in the open state and simulated for 1
μs. The minimum distances between the MTSEA+ mutated residues (I325, T329, A359, I360)
and the PUFA head groups showed significant interactions varying between 2-4 Å (Figure
III.8 B). To further verify this result, I included a negative control also originating from
experimental data; mutating the L366 residue did not induce PUFA-mediated shifts in
Shaker activation [129]. Indeed, the minimum distance between the PUFA head group and
L366 remained larger than 10 Å during 500 ns of simulation, which is in stark contrast to the
PUFA-sensitive mutations (Figure III.8 B).
Chapter H. Study III
81
Figure III.8 Characterization of PUFA interaction sites by introduction of MTSEA+ and
cysteine probes. (A) Side-view of PUFAs interacting with the VSD and a PUFA iso-density
surface at 14 % occupancy depicted in brown mesh. (B) Average minimum distances
between MTSEA+-modified residues and PUFA carboxyl head groups (blue) and carbon tails
(green). (C) Average minimum distances between the cysteine-mutated residues and PUFA
carboxyl head groups (blue) and PUFA carbon tails (green). The error bars display the
standard errors of the VSD-PUFA minimum distances.
In addition, to differentiate between the role of the PUFA head groups and tails and
their specific interactions with the channel, I also monitored distances between the carbon
tails and the mutated residues. In these analyses, the carbon tails were restricted to
carbons C12-22 in the PUFA carbon chain near the terminal methyl group to enable better
discrimination between PUFA head and tail. In contrast to PUFA head groups, the PUFA
carbon tails maintained a ~1 Å larger minimum distance for all mutations except the I360
mutation which has a slightly lower average distance although a larger standard error in the
head group minimum distance may account for this divergence (Figure III.8 B). A similar
tendency was observed for the control mutation L366, where the PUFA tails displayed a
significantly lower minimum distance to the mutated residue. In summary, the MTSEA +
simulation systems validated the modes of PUFA-protein interactions observed in the WT
system where the PUFA head group interacted specifically to the S3-S4 linker while the
tails maintained less defined protein interactions.
Considering the evident influence on the PUFA-channel interaction caused by
MTSEA+-modified residues, I introduced an additional test to probe whether the effects
Chapter H. Study III
82
remained after the removal of the added charge. At 500 ns, the MTSEA+ modified residues
were mutated back into cysteines and were simulated for an additional 500 ns. Upon
charge removal, a distinct change in the average distances between the cysteine side
chains and the PUFA head groups was observed (Figure III.8 C), with PUFAs on average
displaying a minimum distance of 6 - 10 Å across the four mutated systems. Similarly, the
distances between the PUFA carbon tails and the cysteines were significantly altered,
except for the I325 mutation (Figure III.8 C).
H.4 Discussion
The general effects of PUFAs on both channel activation and inactivation has been
documented for Kv channels [103,105], Nav channels [104,107,237,238], and Cav channels
[107,239], but exactly where these unsaturated fatty acids bind remains less clear. It is also
not ascertained whether it is the ion pore or the VSD that are the targets of action. For
instance, reduction of Kv1.1 channel currents were assigned to PUFA-protein interactions
involving hydrophobic residues lining the cavity of the ion pore in the open state [240]. In
contrast, a mutational study proposed a PUFA-interaction site to be located on the VSD,
specifically the lipid-facing surface of the extracellular side of TM helices S3 and S4 [129].
Atomistic MD simulations are designed to monitor structural dynamics of complex
environments, such as a membrane protein inserted into a lipid environment and solvated
by water and ions. Therefore, the MD simulation approach is highly suitable for identifying
and monitoring PUFA-protein interactions on the molecular level, which would be difficult
to determine experimentally. Because a PUFA-binding site has never been characterized in
atomistic detail, I set out to identify and characterize a potential PUFA binding site. The
observed PUFA interaction site was located on the lipid-facing side of TM helices S3 and S4
on the extracellular side of the Shaker Kv channel in its open state, in agreement with
previous experimental findings [129]. In addition, I observed a significant difference in the
structural dynamics and protein interaction patterns of the PUFA carboxylic head group
and the lipophilic carbon tail. The PUFA head groups favored fewer and highly specific
polar/charged residue interactions along the S3-S4 extracellular part, in contrast to the
carbon tails that interacted non-specifically with a wider range of S3-S4 residues buried in
the hydrophobic core in the lipid bilayer. Furthermore, by monitoring the structural
dynamics displayed by SFAs and PUFAs, I ascertained an increased conformational
flexibility in the polyunsaturated carbon tails compared to saturated carbon chains.
Because the saturation level of fatty acids are highly correlated to modulation of channel
function [128,238,241,242], the structural dynamics of fatty acids in the lipid bilayer are
likely important to obtain optimal interactions to the channel protein. The results indicate
that the flexibility of the unsaturated carbon tail allows the fatty acid to visit several binding
modes that enable a strong specific interaction between the carboxyl head group and the
channel, which would otherwise be inaccessible to a more rigid saturated carbon tail. A
similar conclusion was drawn from recent experimental studies of the Shaker channel [128].
To test this predicted interaction I performed simulations in the presence of SFA molecules.
Indeed, the saturated carbon tails of the SFAs displayed significantly less interactions to the
protein compared to the PUFA tails in the open state of the channel. In the closed state,
both PUFAs and SFAs formed different interaction patterns compared to the open state
Chapter H. Study III
83
simulations. A reduced number of interactions, in particular in the carbon tail interactions,
contrasted the open and closed state interactions.
The increasing flexibility with unsaturation level of fatty acids is well documented.
For instance, NMR studies have shown DHA to undergo rapid conformational transitions
with short correlation times and exceptionally low deuterium order parameters [243,244].
In addition, quantum mechanical calculations have shown that polyunsaturated chains
sample greater conformational space around the cis double bonds with more rapid
reorientations near the methyl end of the chain [233]. The increased conformational
flexibility in PUFAs has been explained by lowered torsional energy barriers for the
rotatable bonds in these carbon chains [245]. In addition, several rhodopsin studies show
PUFA-specific modulation [245,246]. Therefore, tail flexibility enabling PUFA electrostatic
head group interactions may be a general mechanism.
The interactions between the PUFA head groups and the VSD in the open state
were concentrated to a few residues displaying high contact frequencies. While all
interactions did not occur across all four subunits, sampling interactions across subunits for
extended microsecond simulations times enabled identification of the major interaction
residues. Interestingly, one of these residues was R362, which is the first (R1) of the gating
charge residues. Because involvement of gating charges has also been observed in Shaker
Kv experiments [129,247], it is indeed possible that PUFAs affect channel function by
binding to our proposed binding site and reach the R1 gating charge from this position. In a
recent study, mutations of residues M356 and A359 into arginines increased the K +
channel’s sensitivity to PUFAs considerably [247]. These residues, which are positioned on
the S4 helix, also show up in our contact analyses but with lower contact frequencies, which
might reflect limited sampling. In a study by Xu et al., it was established that higher
hydrophobicity in a ten-residue segment in the extracellular part of the S3 helix in the
paddle chimera helps to stabilize the open state of the channel [248]. Out of these ten
residues, residues I325, T326, T329, V331, and A332 were identified as close contacts to
PUFAs in our simulations of the open state channel, indicative of the role PUFAs play in
stabilizing the open state of the channel.
Further comparison of the observed PUFA-protein interaction pattern to that of
other lipid modulators such as the phosphatidylinositol-4,5-biphosphate (PIP2) lipids hints
at significant complexity. Recently, PIP2 lipids were found to stabilize the closed state of
Kv7.1 channels by interacting with the lower residues of S4. However, in an open state PIP2
lipids migrated to the pore domain (PD) to form salt bridges with the S6 terminus and in
this way meditated coupling between the VSD and the PD [249]. PIP2 lipids were also
observed to migrate from the S4-S5 linker to the S2-S3 linker in KCNQ2 Kv channels to
control deactivation kinetics [250]. Thus, MD simulations have identified putative structural
features underlying the function of modulatory agents that appear fundamentally
different. Future experimental efforts are needed to verify the interaction patterns
presented in our study. Finally, the simulations studies provide a structural framework for
future studies aimed at determining the free energy associated with Kv channel activation
and deactivation in the presence of modulatory agents.
Chapter I. Study IV
84
Chapter I
Chapter I. Study IV
85
Study IV
I.1 The Regulatory Role of Polyunsaturated Fatty
Acids on the Free-Energy Landscape of KV Shaker
Channel Deactivation
High-resolution X-ray structures of KV channels, in particular that of the chimera channel
KV2.1/KV1.2 [82], have been instrumental in the progress of gaining an improved
understanding of the structure and function of voltage-gated ion channels. In moving
towards a more refined picture of the mechanisms of voltage sensing and channel
activation recent structural models together with MD simulations [221,235,251,252], as well
functional characterizations [253,254] have been major contributors. There are three
different functional and conformational states of the voltage sensor that have been
proposed by experimental studies [255]; a polarized membrane causing the resting closed
state (C), an active open state (O) due to depolarization of the membrane, and an openinactivated channel that results from prolonged depolarization. Free energy barriers are
what separate these three states [256], and through applying an external potential the
relative free energy of each state changes, causing the channel to transition into a different
state [257]. One should bear in mind that in effect all experimentally solved structures of
the voltage sensor are of the open state conformation of the channel and all models of
every other conformational state are theoretical models that require experimental
validation.
Channel activation involves the independent outward movement of the four S4
helices, of which three models have been proposed that explain the exact movement of this
Chapter I. Study IV
86
helix and its relation to the other helices in the VSD. The sliding helix model [94,95]
suggests that the positively charged (basic) residues on S4 make contact with counter
negatively charged (acidic) residues on neighboring parts of the VSD by a substantial
translational and rotational movement. The transporter-like model posits a large rotational
but lacking in a significant translational movement of S4 in transferring charges from the
intracellular to the extracellular matrix [96,97]. The third model that followed the
publication of the crystal structure of KvAP [97] is known as the paddle model, which
proposes the conjoint movement of S4 and S3, known as the voltage sensor paddle, during
gating which reach a transmembrane position in the activated state.
With S4 moving down in sequential steps in the gating process, the existence of
metastable intermediate states has been suggested by extended MD simulations that are
also supported by experimental data [220,221]. In each state, one charged residue on S4
passes a barrier, with four separate energy barriers to overcome. The activation and
deactivation barriers are possibly generated by a hydrophobic zone in the VSD that each
basic residue on S4 crosses during the gating process [258]. The well-conserved aromatic
residue, F290, located in the middle of the S2 helix, was first suggested by Long et al. to
play the role of the hydrophobic plug (Figure IV.1). Shown to be important for gating, F290
has been observed to interact with all four gating charges on S4 [259,260].
Figure IV.1 Voltage-sensor domain comparison. The voltage-sensor domain of the Shaker
KV channel in its open (white) state is superimposed on its putative intermediary C1 state
Chapter I. Study IV
87
(blue), depicting the translation of R4 along S4 past the hydrophobic plug F290 located on
helix S2.
The idea of secondary structure changes in S4 during different stages of gating has
been critical in explaining stabilization during gating. Supported by crystal structures and
experiments including histidine scans and MD simulations, a 10-residue stretch on S4,
downstream of R3, adopts a 310-helix conformation in the resting and active states, but
transitions into an -helix when entering into the open-inactivated state [82,255,261-266].
By incorporating a 310-helix segment, with three residues per turn, S4 becomes more
elongated and tightly wound, and as a consequence the arginine side chains align on the
same side on the helix and stabilize the VSD by their interactions with the basic residues on
S2 and S3 during gating and inhibit large distortions in the VSD [94,95,255,264,267].
Previous gating measurement and mutational studies have shown that PUFAs
mainly affect the coordinated opening step of a KV channel and have less impact on the
earlier stages of transition [128,129]. By affecting the final S4 transition, which is closely
linked to the opening of the channel, the larger PUFA effect can be explained by the
outward movement of the gating charges moving them closer to the putative PUFA
binding site on the VSD [129].
In this study, I compute the energetics of the free-energy landscape of the Shaker KV
channel in lipid bilayers free from as well as enriched with PUFAs, using an enhanced
sampling approach, namely accelerated-weight histogram (AWH). By choosing a reaction
coordinate along the vertical translation of S4 towards its down state in the deactivation
pathway, we investigate the free energy differences in passing the first energy barrier of
deactivation, i.e. going from the open (O) state to the closed-1 (C1) state in a KV channel
that remains either affected or unaffected by PUFAs.
I.2 Methods
I.2.1 Shaker channel in PUFA and PUFA-free systems
The simulation model of the tetrameric KV Shaker channel was adopted from the previously
modeled (see Study III) open channel configuration of the channel based on the highresolution X-ray structure of the KV2.1 paddle–KV1.2 chimera channel (PDB ID 2R9R) [82].
The Shaker system in this study was taken from the fully equilibrated 1 μs simulation of wild-type Shaker inserted into a POPC membrane. Likewise, the PUFA-Shaker setup was
taken from the MTSEA+ modified Shaker channel at residue I325 embedded in a PUFA
enriched membrane bilayer simulated for 1 μs. Experimental data have shown how the
voltage-dependence of the Shaker channel shifts by the introduction of a positively charged
MTSEA+ probe at this site [129], results which were corroborated in the previous study
(Study III). For this reason, a mutated Shaker channel was selected for the PUFA-Shaker
system as this mutation has been observed to directly modulate function by interactions to
PUFA molecules. The simulation systems were set up as described previously in Study III
and carried out by Gromacs [170,171] using the CHARMM36 force field [231] with 5 fs time
steps.
Chapter I. Study IV
88
I.2.2 Accelerated weight histogram calculations
In a study by Henrion et al. [235], several intermediate closed states (C1, C2, C3) between the
open (O) and resting states of the Shaker channel were modeled based on a set of
experimental constraints, mainly the formation of metal-ion bridges between specific
residues alongside helices S3 and S4 in the VSD allowed the generation of Rosetta models
which were verified by molecular simulations. The structural information from the
generated intermediate closed-1 state (C1) permitted investigation of free energy
differences between activation and the first step of deactivation (O to C1) in the Shaker KV
channel in the absence and presence of PUFA in the membrane bilayer. The accelerated
weight histogram (AWH) method was used to sample the O-C1 transition by applying a
harmonic potential or umbrellas to the center-of-mass of a pull group on helix S4 and the
center-of-mass of a reference group made up of helices S1-S3 [145]. The pull group
incorporated the backbone residues 365-375 on S4 and the charged side chains of R2-R3R4-K5, which were given a relative weight of 2.7. AWH is an extended ensemble sampling
technique, which adaptively biases the simulation to promote exploration of the free
energy landscape by using a probability weight histogram that allows for efficient free
energy updates. With a pull direction along an updated vector defined by the backbone
residues of helices S1 and S3, the AWH simulations started from the open state (O-state)
pulling down towards the C1 state (C1-state). As one of perks of the AWH method is
repeated passage over the reaction coordinate, transitions from the O to C1 state allowed
for exploration of multiple pathways. Diffusion constant along reaction coordinate was set
to 1e-6 (nm2/ps), and a maximum and a minimum reaction coordinate were fixed with an
interval of 8 Å. The AWH simulation for the Shaker and the PUFA-Shaker systems was run
for 5 μs each.
I.3 Results
I.3.1 Sampling approach
With a number of different ways of computing free energy, the main complication lies in
extracting accurate free energies from molecular simulations. The thorough exploration of
free energy landscapes requires sampling large conformational changes and slow
transitions. I, therefore, have chosen the accelerated weight histogram (AWH) method to
calculate free energy along a reaction coordinate [145]. A technique that by using a
probability weight histogram enables efficient free energy updates, AWH enhances
sampling by adaptively biasing the simulation to promote exploration of the free energy
landscape. A proper choice of reaction coordinate increases the scope of exploration of the
configuration space and consequently resulting in a more accurate estimate of the freeenergy profile along the chosen reaction coordinate. In the event of a poorly chosen
reaction coordinate, detection is facilitated by the AWH method, as the weight histogram
would in such cases deviate significantly from the target distribution.
I have calculated the energetics of the O to C1 deactivation landscape of the Shaker
KV channel in its atomistic description inserted in both PUFA-free and PUFA-enriched lipid
Chapter I. Study IV
89
bilayers, using the AWH method. In selecting a suitable reaction coordinate to enable an
enhanced sampling of the deactivation pathway, a natural choice is one associated with the
deactivation pathway itself, namely, the vertical displacement of helix S4. In this
framework, the reaction coordinate is determined by the spatial translation of the gating
charge R4 located on S4 in transitioning from the O state to the C1 state of the channel,
progressing down the deactivation pathway along a vector defined by helices S1 and S3 of
the VSD (Figure IV.1). The free energies associated with the presence and absence of
PUFAs on the Shaker KV channel were investigated in 5 μs long AWH samplings. However,
as transitions along the deactivation pathway slowed down significantly and entire O - C1
transitions barely took place beyond 1 μs into the simulations, I decided to focus on this
initial part of the sampling data.
I.3.2 PUFA affects the free-energy landscape of deactivation
Sampling over the course of 1 μs allowed the calculation of free energies of deactivation for the Shaker KV channel in membrane bilayers both free of and enriched with PUFAs. The
resulting free-energy surface reveals striking differences in favorable states along the
deactivation pathway (Figure IV.2). The average free energy, calculated across the four
subunits of the channel, reveals a lower free energy in the open state of the K V channel in
the PUFA enriched system with a difference of ~ 8 kcal/mol. In contrast, at the intermediary
C1 state of the channel, with an energy level ~ 10 kcal/mol lower, the K V channel in the
PUFA free membrane is at a more stable state. Noteworthy, is the comparison of the
energy landscape before and after the transitioning across the hydrophobic barrier. Prior to
the transition across this barrier, the PUFA enriched system maintains a much lower free
energy, however, once the barrier has been crossed and the transition to the metastable C1
state is approaching the PUFA enriched system enters into a higher free-energy state.
Chapter I. Study IV
90
Figure IV.2 Free-energy profile of the O to C1 transition. Free-energy profiles along S4
backbone displacement for the Shaker KV channel in PUFA free (blue) and PUFA enriched
(green) lipid bilayers. The error bars display the standard error of free energy across the
four channel subunits. The shaded area depicts the energy barrier characterized by a
hydrophobic zone along S2 and S4.
The large standard errors and the shape of the free energy profile lacking a barrier
distinguishing the activated state from the intermediate C1 state by the absence of two
stable minima suggests that a much more thorough sampling of the deactivation pathway
may be required. This could increase the possibility that a finer description of the
configurational space can potentially give rise to a smoother free-energy landscape with a
visible barrier separating the two thermodynamically stable states. Deterred by
overcoming the hydrophobic barrier within the scope of current sampling, I deduce that the
obtained free-energy profile has not converged and requires further work.
I.3.3 PUFA stabilizes the open state configuration
The distortion of the channel during the transitions can be assessed by calculating the root
mean-square deviation (RMSD) of the entire VSD and segments S1-S3. The tendency of the
PUFA enriched system to stabilize the open state of the Shaker KV channel as observed by
the free energy landscape, is also reflected in RMSD measurements of the VSD of the
channel as illustrated in Figure IV.3. This is particularly evident in the RMSD of the initial O
to C1 transition. With the S1-S3 domain of the KV channel in the PUFA enriched membrane
bilayer deviating considerably less from the open state crystal structure, an RMSD between
1.5 – 2.0 Å is observed. In contrast, the S1-S3 domain of the PUFA free system displays
higher perturbation and initially has an RMSD of roughly 2.0 – 2.5 Å. However, following
Chapter I. Study IV
91
the initial O to C1 transition in which S4 has translated down once, both the PUFA free and
PUFA enriched systems converge to comparable RMSD values, fluctuating around an
RMSD of ~ 2 Å. A similar pattern is observed in the RMSD for the entire VSD of the channel
in both systems in which the same behavior is observed except for slightly higher RMSD
values, which is due to the inclusion of S4 that contributes further by the pull-down motion
along the deactivation pathway (Figure IV.3, inset).
Figure IV.3 Root-mean square deviation (RMSD) of the backbone atoms of segments S1-S3
during 1 μs of AWH sampling. S4 is translated in a downward motion driving the channel from an open state to an intermediate C1 state in PUFA free (blue) versus PUFA enriched
(green) systems. RMSD was calculated relative to the initial conformation after a least
squares fitting of S1–S3, with error bars indicating the standard errors across the four
subunits of the channel. (Inset) RMSD for the entire VSD compared in both systems.
I.3.4 310-helix integrity
To further investigate the structural dynamics of S4 during translation into a down state, I
performed secondary structure evolution calculations to obtain a clearer picture of
structural element formations of the S4 helix when the channel transitions into the C1 state.
As illustrated in Figure IV.4 A, in both systems free of and enriched with PUFAs, the S4
residues between R1 and K5, largely adopt a 310-helix formation during the initial transition.
It is beyond this first transition that a differing trend separates the two systems. Beyond the
Chapter I. Study IV
92
initial transition frame, there is a spontaneous change of secondary structure in S4, with
the 310-helix region losing its stability and converting to an -helix in the PUFA enriched
system.
Figure IV.4 Secondary structure evolution of S4 segment. 310-helix formation in the R1-K5
region evolves as S4 translates from the O state down towards the C 1 state along the
Chapter I. Study IV
93
reaction coordinate in the PUFA free (left panel) and PUFA enriched (right panel) systems,
displayed individually for each subunit of the tetrameric channel.
In general, the 310-helix propensity for the R1-K5 segment remains higher for the PUFA free
system as opposed to the PUFA enriched system (Figure IV.5). Although this can be
interpreted as support for stabilization of a 310 conformation as S4 translates down in a
PUFA free system, it is worth noting that this could be due to totally arbitrary events as the
derived free-energy landscape for both PUFA free and enriched systems lacked proper
energy barriers and could not be totally relied upon.
Figure IV.5 310-helix propensity. Probability distribution of 310-helix formation for a 21residue segment on S4 in systems free of (orange) and enriched with (maroon) PUFAs.
I.4 Discussion
In recent years, literature investigating lipid-protein interactions have increasingly been
highlighting the fundamental importance of lipids on voltage sensor function. Both
experimental and simulation studies (see Study III) have generated substantial evidence
Chapter I. Study IV
94
towards PUFAs modulating KV channel function by interacting with a specific binding site
situated at the protein-lipid interface of the VSD in proximity of the extracellular halves of
the S3 and S4 helices, thereby stabilizing the open active state of the channel [129]. In
particular, gating measurement studies have identified the last step of transition towards
an open state in the gating pathway to be dominated by the PUFA effect [128,129].
Naturally, a relevant and interesting question, which follows, is the estimation of the free
energy difference that arises from the effects of PUFAs in transitioning one step down the
deactivation pathway, from the open state to the C1 state.
These results, despite its’ limitations, has shed light on apparent free energy
differences during the O – C1 transition. The PUFAs’ close interactions with the KV channel
results in a lower free energy at the open state stabilizing this conformation of the channel.
As the PUFAs bring the channel into a thermodynamically more favorable state, their
absence on the other hand, results in a relatively higher free energy enabling the transition
to the C1 state more readily. Once past the hydrophobic barrier, the absence of the effect of
PUFAs on the channel leads to lower free energy compared to an energetically more
unstable C1 conformational state of the channel that is interacting with PUFAs. Although it
was not possible to pull slowly enough to obtain repeated O - C1 transitions to derive
statistically accurate work profiles, results from our initial simulations indicate that further
sampling of the deactivation pathway is required. As previously mentioned, the large
standard errors in addition to the non-existence of clear barriers separating the two
energetically stable states, the open versus the C1 intermediate, are indicative of sampling
that has not converged and demands additional exploration of the free energy landscape
along the deactivation pathway. In fact, in order to ensure continued and consistent PUFAchannel interaction the introduction of a single mutation to the channel may very well have
had its effects on the statistical sampling and contributed to inaccuracies on the free
energy landscape of the VSD and altered the intrinsic equilibrium between the states.
The conformational change of an -helix into a 310-helix of the S4 segment that
carries the gating charges has been linked as a component of the inactivation barrier
[255,256,264,266]. The 310-helix adoption has been suggested to lower the energetic cost
of the intermediate barriers, thus stabilizing the -helix after crossing the hydrophobic
barrier requires higher energies than that of the 310-helix. This explains the higher free
energy associated with the PUFA enriched system after passing the hydrophobic barrier in
this study, in which S4 fails to keep its 310 helical conformation beyond the initial transition
to the C1 state.
Concluding Remarks and Future Perspectives
95
Concluding Remarks and Future Perspectives
Every question answered in biology corresponds to the correct placement of a piece of a
jigsaw puzzle. With every piece, as one part of the complete picture is revealed, hints are
given as to the remaining questions. A theoretical approach in solving the jigsaw puzzle has
been tremendously instrumental. However, the application of computational methods in
exploring the dynamics of biological systems alongside its’ strengths carries limitations as well. The strengths and limitations have naturally been reflected in the different studies
carried out as part of this thesis.
The strengths of computational methods particularly manifest themselves when wet-lab
experiments are unable to capture certain states of structures or the relevant dynamics
associated with protein function. The structural characterization of membrane proteins are
for instance challenging, as the requirement of a membrane environment for purification
and crystallization is tricky to fulfill or solving the structure of a highly dynamic and
potentially disordered N-terminal extension of water-soluble proteins as witnessed in this
thesis. Capturing a dynamical process such as the gating mechanism of ion channels or the
structural effects of post-translational modifications of proteins is not easily achieved by
experimental procedures. Modeling techniques as well as simulations have enabled the
study of these processes. We have obtained a clearer picture of the structural elements of
IκBα and the effects of phosphorylation on the dynamic properties of this transcription
factor inhibitor. Simulations have shed light on the dynamic nature of interactions of
PUFAs with the Shaker KV channel and aided in understanding the underlying mechanisms.
The limitations of computational methods are equally many. Limited computational power
leading to poor covering of biologically relevant time scales and thorough exploration of
the entire configurational space with non-optimal sampling methods is one of the major
drawbacks. Improved force fields with parameter validations are also necessary for
accurate simulations. Inefficient and insufficient sampling was especially evident during the
free energy calculations of the first step of deactivation of the Shaker KV channel, and
additional sampling are absolutely required to obtain a more accurate free energy
landscape.
Bibliography
96
Bibliography
1. Lehninger AL, Nelson DL, Cox MM (2008) Lehninger principles of biochemistry. New
York: W.H. Freeman.
2. What Are Proteins? Central Dogma of Molecular Biology.
3. Kendrew JC, Bodo G, Dintzis HM, Parrish RG, Wyckoff H, et al. (1958) A threedimensional model of the myoglobin molecule obtained by x-ray analysis. Nature 181:
662-666.
4. Perutz MF, Rossmann MG, Cullis AF, Muirhead H, Will G, et al. (1960) Structure of
haemoglobin: a three-dimensional Fourier synthesis at 5.5-A. resolution, obtained by Xray analysis. Nature 185: 416-422.
5. Ramachandran GN, Ramakrishnan C, Sasisekharan V (1963) Stereochemistry of
polypeptide chain configurations. J Mol Biol 7: 95-99.
6. Ramachandran GN, Sasisekharan V (1968) Conformation of polypeptides and proteins.
Adv Protein Chem 23: 283-438.
7. Pauling L, Corey RB (1950) Two Hydrogen-Bonded Spiral Configurations of the
Polypeptide Chain. Journal of the American Chemical Society 72: 5349-5349.
8. Pauling L, Corey RB (1951) The pleated sheet, a new layer configuration of polypeptide
chains. Proc Natl Acad Sci U S A 37: 251-256.
9. Pauling L, Corey RB, Branson HR (1951) The structure of proteins: Two hydrogenbonded helical configurations of the polypeptide chain. Proceedings of the National
Academy of Sciences 37: 205-211.
10. Protein Structure I.
11. Yang AS, Honig B (1995) Free energy determinants of secondary structure formation: I.
alpha-Helices. J Mol Biol 252: 351-365.
12. Ermolenko DN, Richardson JM, Makhatadze GI (2003) Noncharged amino acid residues
at the solvent-exposed positions in the middle and at the C terminus of the alpha-helix
have the same helical propensity. Protein Science 12: 1169-1176.
13. Vieira-Pires RS, Morais-Cabral JH (2010) 3(10) helices in channels and other membrane
proteins. Journal of General Physiology 136: 585-592.
14. Farzadfard F, Gharaei N, Pezeshk H, Marashi SA (2008) Beta-sheet capping: signals that
initiate and terminate beta-sheet formation. J Struct Biol 161: 101-110.
15. Rose GD (1978) Prediction of chain turns in globular proteins on a hydrophobic basis.
Nature 272: 586-590.
Bibliography
97
16. Fersht AR, Matouschek A, Serrano L (1992) The folding of an enzyme. I. Theory of
protein engineering analysis of stability and pathway of protein folding. J Mol Biol 224:
771-782.
17. Chothia C, Janin J (1975) Principles of protein-protein recognition. Nature 256: 705-708.
18. Groves MR, Barford D (1999) Topological characteristics of helical repeat proteins. Curr
Opin Struct Biol 9: 383-389.
19. Kobe B, Kajava AV (2000) When protein folding is simplified to protein coiling: the
continuum of solenoid protein structures. Trends Biochem Sci 25: 509-515.
20. Main ER, Jackson SE, Regan L (2003) The folding and design of repeat proteins:
reaching a consensus. Curr Opin Struct Biol 13: 482-489.
21. Breeden L, Nasmyth K (1987) Similarity between cell-cycle genes of budding yeast and
fission yeast and the Notch gene of Drosophila. Nature 329: 651-654.
22. Lux SE, John KM, Bennett V (1990) Analysis of cDNA for human erythrocyte ankyrin
indicates a repeated structure with homology to tissue-differentiation and cell-cycle
control proteins. Nature 344: 36-42.
23. Marcotte EM, Pellegrini M, Yeates TO, Eisenberg D (1999) A census of protein repeats.
Journal of Molecular Biology 293: 151-160.
24. Mosavi LK, Cammett TJ, Desrosiers DC, Peng ZY (2004) The ankyrin repeat as
molecular architecture for protein recognition. Protein Science 13: 1435-1448.
25. Sedgwick SG, Smerdon SJ (1999) The ankyrin repeat: a diversity of interactions on a
common structural framework. Trends Biochem Sci 24: 311-316.
26. Sen R, Baltimore D (1986) Inducibility of kappa immunoglobulin enhancer-binding
protein Nf-kappa B by a posttranslational mechanism. Cell 47: 921-928.
27. Sen R, Baltimore D (1986) Multiple nuclear factors interact with the immunoglobulin
enhancer sequences. Cell 46: 705-716.
28. Baldwin AS, Jr. (1996) The NF-kappa B and I kappa B proteins: new discoveries and
insights. Annual Review of Immunology 14: 649-683.
29. Bonizzi G, Karin M (2004) The two NF-kappaB activation pathways and their role in
innate and adaptive immunity. Trends Immunol 25: 280-288.
30. Ghosh S, May MJ, Kopp EB (1998) NF-kappa B and Rel proteins: evolutionarily
conserved mediators of immune responses. Annual Review of Immunology 16: 225-260.
31. Miyamoto S, Verma IM (1995) Rel/NF-kappa B/I kappa B story. Adv Cancer Res 66:
255-292.
32. Attar RM, Caamano J, Carrasco D, Iotsova V, Ishikawa H, et al. (1997) Genetic
approaches to study Rel/NF-kappa B/I kappa B function in mice. Semin Cancer Biol 8:
93-101.
Bibliography
98
33. Whiteside ST, Israel A (1997) I kappa B proteins: structure, function and regulation.
Semin Cancer Biol 8: 75-82.
34. Vallabhapurapu S, Karin M (2009) Regulation and function of NF-kappaB transcription
factors in the immune system. Annual Review of Immunology 27: 693-733.
35. Karin M, Ben-Neriah Y (2000) Phosphorylation meets ubiquitination: The control of NFkappa B activity. Annual Review of Immunology 18: 621-+.
36. Arenzana-Seisdedos F, Turpin P, Rodriguez M, Thomas D, Hay RT, et al. (1997) Nuclear
localization of I kappa B alpha promotes active transport of NF-kappa B from the
nucleus to the cytoplasm. J Cell Sci 110 ( Pt 3): 369-378.
37. Hoffmann A, Levchenko A, Scott ML, Baltimore D (2002) The IkappaB-NF-kappaB
signaling module: temporal control and selective gene activation. Science 298: 12411245.
38. Chen L, Fischle W, Verdin E, Greene WC (2001) Duration of nuclear NF-kappaB action
regulated by reversible acetylation. Science 293: 1653-1657.
39. DiDonato JA, Hayakawa M, Rothwarf DM, Zandi E, Karin M (1997) A cytokineresponsive IkappaB kinase that activates the transcription factor NF-kappaB. Nature 388:
548-554.
40. Mercurio F, Zhu H, Murray BW, Shevchenko A, Bennett BL, et al. (1997) IKK-1 and
IKK-2: cytokine-activated IkappaB kinases essential for NF-kappaB activation. Science
278: 860-866.
41. Woronicz JD, Gao X, Cao Z, Rothe M, Goeddel DV (1997) IkappaB kinase-beta: NFkappaB activation and complex formation with IkappaB kinase-alpha and NIK. Science
278: 866-869.
42. Yamaoka S, Courtois G, Bessia C, Whiteside ST, Weil R, et al. (1998) Complementation
cloning of NEMO, a component of the IkappaB kinase complex essential for NF-kappaB
activation. Cell 93: 1231-1240.
43. Mercurio F, Murray BW, Shevchenko A, Bennett BL, Young DB, et al. (1999) IkappaB
kinase (IKK)-associated protein 1, a common component of the heterogeneous IKK
complex. Mol Cell Biol 19: 1526-1538.
44. Huynh QK, Boddupalli H, Rouw SA, Koboldt CM, Hall T, et al. (2000) Characterization
of the recombinant IKK1/IKK2 heterodimer. Mechanisms regulating kinase activity.
Journal of Biological Chemistry 275: 25883-25891.
45. Beg AA, Sha WC, Bronson RT, Ghosh S, Baltimore D (1995) Embryonic lethality and
liver degeneration in mice lacking the RelA component of NF-kappa B. Nature 376: 167170.
46. Delhase M, Hayakawa M, Chen Y, Karin M (1999) Positive and negative regulation of
IkappaB kinase activity through IKKbeta subunit phosphorylation. Science 284: 309313.
Bibliography
99
47. Brown K, Gerstberger S, Carlson L, Franzoso G, Siebenlist U (1995) Control of I kappa
B-alpha proteolysis by site-specific, signal-induced phosphorylation. Science 267: 14851488.
48. Karin M (1999) The beginning of the end: IkappaB kinase (IKK) and NF-kappaB
activation. Journal of Biological Chemistry 274: 27339-27342.
49. Brasier AR (2006) The NF-kappaB regulatory network. Cardiovasc Toxicol 6: 111-130.
50. Brasier AR, Jamaluddin M, Casola A, Duan W, Shen Q, et al. (1998) A promoter
recruitment mechanism for tumor necrosis factor-alpha-induced interleukin-8
transcription in type II pulmonary epithelial cells. Dependence on nuclear abundance of
Rel A, NF-kappaB1, and c-Rel transcription factors. Journal of Biological Chemistry
273: 3551-3561.
51. Kaltschmidt B, Kaltschmidt C (2009) NF-kappaB in the nervous system. Cold Spring
Harb Perspect Biol 1: a001271.
52. Karin M (2006) Nuclear factor-kappaB in cancer development and progression. Nature
441: 431-436.
53. Masters SL, Simon A, Aksentijevich I, Kastner DL (2009) Horror autoinflammaticus: the
molecular pathophysiology of autoinflammatory disease (*). Annu Rev Immunol 27:
621-668.
54. Henkel T, Machleidt T, Alkalay I, Kronke M, Ben-Neriah Y, et al. (1993) Rapid
proteolysis of I kappa B-alpha is necessary for activation of transcription factor NFkappa B. Nature 365: 182-185.
55. Deshaies RJ (1999) SCF and Cullin/Ring H2-based ubiquitin ligases. Annu Rev Cell Dev
Biol 15: 435-467.
56. Wu G, Xu G, Schulman BA, Jeffrey PD, Harper JW, et al. (2003) Structure of a betaTrCP1-Skp1-beta-catenin complex: destruction motif binding and lysine specificity of
the SCF(beta-TrCP1) ubiquitin ligase. Mol Cell 11: 1445-1456.
57. Yaron A, Hatzubai A, Davis M, Lavon I, Amit S, et al. (1998) Identification of the
receptor component of the IkappaBalpha-ubiquitin ligase. Nature 396: 590-594.
58. Dyson HJ, Komives EA (2012) Role of disorder in IκB–NFκB interaction. IUBMB
Life 64: 499-505.
59. Cuff JA, Clamp ME, Siddiqui AS, Finlay M, Barton GJ (1998) JPred: a consensus
secondary structure prediction server. Bioinformatics 14: 892-893.
60. Palopoli L, Rombo SE, Terracina G, Tradigo G, Veltri P (2009) Improving protein
secondary structure predictions by prediction fusion. Information Fusion 10: 217-232.
61. Huxford T, Huang DB, Malek S, Ghosh G (1998) The crystal structure of the
IkappaBalpha/NF-kappaB complex reveals mechanisms of NF-kappaB inactivation. Cell
95: 759-770.
Bibliography
100
62. Jacobs MD, Harrison SC (1998) Structure of an I kappa B alpha/NF-kappa B complex.
Cell 95: 749-758.
63. Croy CH, Bergqvist S, Huxford T, Ghosh G, Komives EA (2004) Biophysical
characterization of the free I kappa B alpha ankyrin repeat domain in solution. Protein
Science 13: 1767-1777.
64. Truhlar SM, Torpey JW, Komives EA (2006) Regions of IkappaBalpha that are critical
for its inhibition of NF-kappaB.DNA interaction fold upon binding to NF-kappaB. Proc
Natl Acad Sci U S A 103: 18951-18956.
65. Cervantes CF, Markwick PRL, Sue SC, McCammon JA, Dyson HJ, et al. (2009)
Functional Dynamics of the Folded Ankyrin Repeats of I kappa B alpha Revealed by
Nuclear Magnetic Resonance. Biochemistry 48: 8023-8031.
66. Schuster M, Annemann M, Plaza-Sirvent C, Schmitz I (2013) Atypical I kappa B proteins
- nuclear modulators of NF-kappa B signaling. Cell Communication and Signaling 11.
67. Ferreiro DU, Cervantes CF, Truhlar SM, Cho SS, Wolynes PG, et al. (2007) Stabilizing
IkappaBalpha by "consensus" design. Journal of Molecular Biology 365: 1201-1216.
68. Pons J, Evrard-Todeschi N, Bertho G, Gharbi-Benarous J, Sonois V, et al. (2007)
Structural studies on 24P-I kappa B alpha peptide derived from a human I kappa B-alpha
protein related to the inhibition of the activity of the transcription factor NF-kappa B.
Biochemistry 46: 2958-2972.
69. Luckey M (2008) Membrane structural biology : with biochemical and biophysical
foundations. Cambridge ; New York: Cambridge University Press. xi, 332 p. p.
70. Hodgkin AL, Huxley AF (1952) A quantitative description of membrane current and its
application to conduction and excitation in nerve. J Physiol 117: 500-544.
71. Purves D (2012) Neuroscience. Sunderland, Mass.: Sinauer Associates.
72. Barnett MW, Larkman PM (2007) The action potential. Pract Neurol 7: 192-197.
73. Harmar AJ, Hills RA, Rosser EM, Jones M, Buneman OP, et al. (2009) IUPHAR-DB: the
IUPHAR database of G protein-coupled receptors and ion channels. Nucleic Acids Res
37: D680-685.
74. Wulff H, Castle NA, Pardo LA (2009) Voltage-gated potassium channels as therapeutic
targets. Nat Rev Drug Discov 8: 982-1001.
75. Gutman GA, Chandy KG, Grissmer S, Lazdunski M, McKinnon D, et al. (2005)
International Union of Pharmacology. LIII. Nomenclature and molecular relationships of
voltage-gated potassium channels. Pharmacol Rev 57: 473-508.
76. Papazian DM, Schwarz TL, Tempel BL, Jan YN, Jan LY (1987) Cloning of genomic and
complementary DNA from Shaker, a putative potassium channel gene from Drosophila.
Science 237: 749-753.
Bibliography
101
77. Doyle DA, Morais Cabral J, Pfuetzner RA, Kuo A, Gulbis JM, et al. (1998) The structure
of the potassium channel: molecular basis of K+ conduction and selectivity. Science 280:
69-77.
78. Heginbotham L, Lu Z, Abramson T, MacKinnon R (1994) Mutations in the K+ channel
signature sequence. Biophys J 66: 1061-1067.
79. Gulbis JM, Zhou M, Mann S, MacKinnon R (2000) Structure of the cytoplasmic beta
subunit-T1 assembly of voltage-dependent K+ channels. Science 289: 123-127.
80. Kreusch A, Pfaffinger PJ, Stevens CF, Choe S (1998) Crystal structure of the
tetramerization domain of the Shaker potassium channel. Nature 392: 945-948.
81. Shen NV, Pfaffinger PJ (1995) Molecular recognition and assembly sequences involved
in the subfamily-specific assembly of voltage-gated K+ channel subunit proteins. Neuron
14: 625-633.
82. Long SB, Tao X, Campbell EB, MacKinnon R (2007) Atomic structure of a voltagedependent K+ channel in a lipid membrane-like environment. Nature 450: 376-382.
83. Aggarwal SK, MacKinnon R (1996) Contribution of the S4 segment to gating charge in
the Shaker K+ channel. Neuron 16: 1169-1177.
84. Seoh SA, Sigg D, Papazian DM, Bezanilla F (1996) Voltage-sensing residues in the S2
and S4 segments of the Shaker K+ channel. Neuron 16: 1159-1167.
85. Jogini V, Roux B (2007) Dynamics of the Kv1.2 voltage-gated K(+) channel in a
membrane environment. Biophysical Journal 93: 3070-3082.
86. Schmidt D, Jiang QX, MacKinnon R (2006) Phospholipids and the origin of cationic
gating charges in voltage sensors. Nature 444: 775-779.
87. Shafrir Y, Durell SR, Guy HR (2008) Models of voltage-dependent conformational
changes in NaChBac channels. Biophys J 95: 3663-3676.
88. Keynes RD, Elinder F (1998) Modelling the activation, opening, inactivation and
reopening of the voltage-gated sodium channel. Proc Biol Sci 265: 263-270.
89. Ledwell JL, Aldrich RW (1999) Mutations in the S4 region isolate the final voltagedependent cooperative step in potassium channel activation. J Gen Physiol 113: 389-414.
90. Schoppa NE, Sigworth FJ (1998) Activation of Shaker potassium channels. III. An
activation gating model for wild-type and V2 mutant channels. J Gen Physiol 111: 313342.
91. Sigg D, Stefani E, Bezanilla F (1994) Gating current noise produced by elementary
transitions in Shaker potassium channels. Science 264: 578-582.
92. Zagotta WN, Hoshi T, Aldrich RW (1994) Shaker potassium channel gating. III:
Evaluation of kinetic models for activation. J Gen Physiol 103: 321-362.
Bibliography
102
93. Borjesson SI, Elinder F (2008) Structure, function, and modification of the voltage sensor
in voltage-gated ion channels. Cell Biochem Biophys 52: 149-174.
94. Catterall WA (1986) Voltage-dependent gating of sodium channels: correlating structure
and function. Trends in Neurosciences 9: 7-10.
95. Guy HR, Seetharamulu P (1986) Molecular model of the action potential sodium channel.
Proc Natl Acad Sci U S A 83: 508-512.
96. Papazian DM, Bezanilla F (1997) How does an ion channel sense voltage? News in
Physiological Sciences 12: 203-210.
97. Jiang YX, Ruta V, Chen JY, Lee A, MacKinnon R (2003) The principle of gating charge
movement in a voltage-dependent K+ channel. Nature 423: 42-48.
98. Elinder F, Arhem P, Larsson HP (2001) Localization of the extracellular end of the
voltage sensor S4 in a potassium channel. Biophysical Journal 80: 1802-1809.
99. Suh BC, Hille B (2005) Regulation of ion channels by phosphatidylinositol 4,5bisphosphate. Curr Opin Neurobiol 15: 370-378.
100. Catterall WA, Cestele S, Yarov-Yarovoy V, Yu FH, Konoki K, et al. (2007) Voltagegated ion channels and gating modifier toxins. Toxicon 49: 124-141.
101. Swartz KJ (2007) Tarantula toxins interacting with voltage sensors in potassium
channels. Toxicon 49: 213-230.
102. Lee HC, Wang JM, Swartz KJ (2003) Interaction between extracellular Hanatoxin and
the resting conformation of the voltage-sensor paddle in Kv channels. Neuron 40: 527536.
103. Honore E, Barhanin J, Attali B, Lesage F, Lazdunski M (1994) External Blockade of the
Major Cardiac Delayed-Rectifier K+ Channel (Kv1.5) by Polyunsaturated Fatty-Acids.
Proceedings of the National Academy of Sciences of the United States of America 91:
1937-1941.
104. Leifert WR, McMurchie EJ, Saint DA (1999) Inhibition of cardiac sodium currents in
adult rat myocytes by n-3 polyunsaturated fatty acids. Journal of Physiology-London
520: 671-679.
105. McKay MC, Worley JF (2001) Linoleic acid both enhances activation and blocks Kv1.5
and Kv2.1 channels by two separate mechanisms. American Journal of Physiology-Cell
Physiology 281: C1277-C1284.
106. Poling JS, Vicini S, Rogawski MA, Salem N, Jr. (1996) Docosahexaenoic acid block of
neuronal voltage-gated K+ channels: subunit selective antagonism by zinc.
Neuropharmacology 35: 969-982.
107. Xiao YF, Sigg DC, Leaf A (2005) The antiarrhythmic effect of n-3 polyunsaturated fatty
acids: Modulation of cardiac ion channels as a potential mechanism. Journal of
Membrane Biology 206: 141-154.
Bibliography
103
108. Xu XP, Erichsen D, Borjesson SI, Dahlin M, Amark P, et al. (2008) Polyunsaturated
fatty acids and cerebrospinal fluid from children on the ketogenic diet open a voltagegated K channel: A putative mechanism of antiseizure action. Epilepsy Research 80: 5766.
109. McNamara JO (1999) Emerging insights into the genesis of epilepsy. Nature 399: A1522.
110. Lefevre F, Aronson N (2000) Ketogenic diet for the treatment of refractory epilepsy in
children: A systematic review of efficacy. Pediatrics 105.
111. Freeman JM, Vining EPG, Pillas DJ, Pyzik PL, Casey JC, et al. (1998) The efficacy of
the ketogenic diet - 1998: A prospective evaluation of intervention in 150 children.
Pediatrics 102: 1358-1363.
112. Vining EPG, Freeman JM, Ballaban-Gil K, Camfield CS, Camfield PR, et al. (1998) A
multicenter study of the efficacy of the ketogenic diet. Archives of Neurology 55: 14331437.
113. Cunnane SC, Musa K, Ryan MA, Whiting S, Fraser DD (2002) Potential role of
polyunsaturates in seizure protection achieved with the ketogenic diet. Prostaglandins
Leukot Essent Fatty Acids 67: 131-135.
114. Spector AA (2001) Plasma free fatty acid and lipoproteins as sources of polyunsaturated
fatty acid for the brain. J Mol Neurosci 16: 159-165; discussion 215-121.
115. Fraser DD, Whiting S, Andrew RD, Macdonald EA, Musa-Veloso K, et al. (2003)
Elevated polyunsaturated fatty acids in blood serum obtained from children on the
ketogenic diet. Neurology 60: 1026-1029.
116. Taha AY, Ryan MA, Cunnane SC (2005) Despite transient ketosis, the classic high-fat
ketogenic diet induces marked changes in fatty acid metabolism in rats. Metabolism 54:
1127-1132.
117. Vreugdenhil M, Bruehl C, Voskuyl RA, Kang JX, Leaf A, et al. (1996) Polyunsaturated
fatty acids modulate sodium and calcium currents in CA1 neurons. Proc Natl Acad Sci U
S A 93: 12559-12563.
118. Xiao Y, Li X (1999) Polyunsaturated fatty acids modify mouse hippocampal neuronal
excitability during excitotoxic or convulsant stimulation. Brain Res 846: 112-121.
119. Biervert C, Schroeder BC, Kubisch C, Berkovic SF, Propping P, et al. (1998) A
potassium channel mutation in neonatal human epilepsy. Science 279: 403-406.
120. Browne DL, Gancher ST, Nutt JG, Brunt ERP, Smith EA, et al. (1994) Episodic Ataxia
Myokymia Syndrome Is Associated with Point Mutations in the Human Potassium
Channel Gene, Kcna1. Nature Genetics 8: 136-140.
121. Cooper EC (2001) Potassium channels: How genetic studies of epileptic syndromes
open paths to new therapeutic targets and drugs. Epilepsia 42: 49-54.
Bibliography
104
122. Eunson LH, Rea R, Zuberi SM, Youroukos S, Panayiotopoulos CP, et al. (2000)
Clinical, genetic, and expression studies of mutations in the potassium channel gene
KCNA1 reveal new phenotypic variability. Annals of Neurology 48: 647-656.
123. Singh NA, Charlier C, Stauffer D, DuPont BR, Leach RJ, et al. (1998) A novel
potassium channel gene, KCNQ2, is mutated in an inherited epilepsy of newborns.
Nature Genetics 18: 25-29.
124. Smart SL, Lopantsev V, Zhang CL, Robbins CA, Wang H, et al. (1998) Deletion of the
K(v)1.1 potassium channel causes epilepsy in mice. Neuron 20: 809-819.
125. Bendahhou S, Cummins TR, Agnew WS (1997) Mechanism of modulation of the
voltage-gated skeletal and cardiac muscle sodium channels by fatty acids. American
Journal of Physiology-Cell Physiology 272: C592-C600.
126. Hallaq H, Smith TW, Leaf A (1992) Modulation of Dihydropyridine-Sensitive Calcium
Channels in Heart-Cells by Fish Oil Fatty-Acids. Proceedings of the National Academy
of Sciences of the United States of America 89: 1760-1764.
127. Kang JX, Leaf A (1996) Evidence that free polyunsaturated fatty acids modify Na+
channels by directly binding to the channel proteins. Proceedings of the National
Academy of Sciences of the United States of America 93: 3542-3546.
128. Borjesson SI, Hammarstrom S, Elinder F (2008) Lipoelectric modification of ion
channel voltage gating by polyunsaturated fatty acids. Biophysical Journal 95: 22422253.
129. Borjesson SI, Elinder F (2011) An electrostatic potassium channel opener targeting the
final voltage sensor transition. Journal of General Physiology 137: 563-577.
130. van der Kamp MW, Shaw KE, Woods CJ, Mulholland AJ (2008) Biomolecular
simulation and modelling: status, progress and prospects. J R Soc Interface 5 Suppl 3:
S173-190.
131. Tai K, Fowler P, Mokrab Y, Stansfeld P, Sansom MS (2008) Molecular modeling and
simulation studies of ion channel structures, dynamics and mechanisms. Methods Cell
Biol 90: 233-265.
132. Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, et al. (1997) Gapped BLAST
and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids
Research 25: 3389-3402.
133. Karplus K, Barrett C, Hughey R (1998) Hidden Markov models for detecting remote
protein homologies. Bioinformatics 14: 846-856.
134. Karchin R, Cline M, Mandel-Gutfreund Y, Karplus K (2003) Hidden Markov models
that use predicted local structure for fold recognition: Alphabets of backbone geometry.
Proteins-Structure Function and Bioinformatics 51: 504-514.
135. Jones DT, Taylor WR, Thornton JM (1992) A New Approach to Protein Fold
Recognition. Nature 358: 86-89.
Bibliography
105
136. Sali A, Blundell TL (1993) Comparative Protein Modeling by Satisfaction of Spatial
Restraints. Journal of Molecular Biology 234: 779-815.
137. Mccammon JA, Gelin BR, Karplus M (1977) Dynamics of Folded Proteins. Nature 267:
585-590.
138. Verlet L (1967) Computer Experiments on Classical Fluids .I. Thermodynamical
Properties of Lennard-Jones Molecules. Physical Review 159: 98-+.
139. Hockney RW (1970) The Potential Calculation and Some Applications. Methods in
Computational Physics 9: 136-211.
140. Hess B, Kutzner C, van der Spoel D, Lindahl E (2008) GROMACS 4: Algorithms for
highly efficient, load-balanced, and scalable molecular simulation. Journal of Chemical
Theory and Computation 4: 435-447.
141. Phillips JC, Braun R, Wang W, Gumbart J, Tajkhorshid E, et al. (2005) Scalable
molecular dynamics with NAMD. Journal of Computational Chemistry 26: 1781-1802.
142. Brooks BR, Brooks CL, Mackerell AD, Nilsson L, Petrella RJ, et al. (2009) CHARMM:
The Biomolecular Simulation Program. Journal of Computational Chemistry 30: 15451614.
143. Kirkwood J (1935) Statistical mechanics of fluid mixtures. Journal of Chemical Physics
3: 300-313.
144. Torrie GM, Valleau JP (1977) Non-Physical Sampling Distributions in Monte-Carlo
Free-Energy Estimation - Umbrella Sampling. Journal of Computational Physics 23:
187-199.
145. Lindahl V, Lidmar J, Hess B (2014) Accelerated weight histogram method for exploring
free energy landscapes. Journal of Chemical Physics 141.
146. Ghosh G, Wang VY-F, Huang D-B, Fusco A (2012) NF-κB regulation: lessons from
structures. Immunological Reviews 246: 36-58.
147. Ferreiro DU, Komives EA (2010) Molecular mechanisms of system control of NFkappaB signaling by IkappaBalpha. Biochemistry 49: 1560-1567.
148. Huxford T, Hoffmann A, Ghosh G (2011) Understanding the Logic of I kappa B:NFkappa B Regulation in Structural Terms. In: Karin M, editor. Nf-Kb in Health and
Disease. pp. 1-24.
149. Jacobs MD, Harrison SC (1998) Structure of an IkappaBalpha/NF-kappaB complex.
Cell 95: 749-758.
150. Traenckner EB, Baeuerle PA (1995) Appearance of apparently ubiquitin-conjugated I
kappa B-alpha during its phosphorylation-induced degradation in intact cells. Journal of
cell science Supplement 19: 79-84.
151. Huxford T, Malek S, Ghosh G (1999) Structure and mechanism in NF-kappa B/I kappa
B signaling. Cold Spring Harbor symposia on quantitative biology 64: 533-540.
Bibliography
106
152. Hatada EN, Nieters A, Wulczyn FG, Naumann M, Meyer R, et al. (1992) The ankyrin
repeat domains of the NF-kappa B precursor p105 and the protooncogene bcl-3 act as
specific inhibitors of NF-kappa B DNA binding. Proc Natl Acad Sci U S A 89: 24892493.
153. Jaffray E, Wood KM, Hay RT (1995) Domain organization of I kappa B alpha and sites
of interaction with NF-kappa B p65. Mol Cell Biol 15: 2166-2172.
154. Sun SC, Elwood J, Greene WC (1996) Both amino- and carboxyl-terminal sequences
within I kappa B alpha regulate its inducible degradation. Molecular and Cellular
Biology 16: 1058-1065.
155. Rost B (2001) Review: Protein secondary structure prediction continues to rise. Journal
of Structural Biology 134: 204-218.
156. Simossis VA, Heringa J (2004) Optimally segmented consensus secondary structure
prediction. SYMPRED server at
157. Rost B, Sander C (1994) CONSERVATION AND PREDICTION OF SOLVENT
ACCESSIBILITY IN PROTEIN FAMILIES. Proteins-Structure Function and Genetics
20: 216-226.
158. Pollastri G, Przybylski D, Rost B, Baldi P (2002) Improving the prediction of protein
secondary structure in three and eight classes using recurrent neural networks and
profiles. Proteins 47: 228-235.
159. Lin K, Simossis VA, Taylor WR, Heringa J (2005) A simple and fast secondary
structure prediction method using hidden neural networks. Bioinformatics 21: 152-159.
160. Jones DT (1999) Protein secondary structure prediction based on position-specific
scoring matrices. Journal of Molecular Biology 292: 195-202.
161. Leman JK, Mueller R, Karakas M, Woetzel N, Meiler J (2013) Simultaneous prediction
of protein secondary structure and transmembrane spans. Proteins: Structure, Function,
and Bioinformatics 81: 1127-1140.
162. Petersen B, Petersen TN, Andersen P, Nielsen M, Lundegaard C (2009) A generic
method for assignment of reliability scores applied to solvent accessibility predictions.
BMC Struct Biol 9: 51.
163. Pollastri G, McLysaght A (2005) Porter: a new, accurate server for protein secondary
structure prediction. Bioinformatics 21: 1719-1720.
164. Rost B, Yachdav G, Liu J (2004) The PredictProtein server. Nucleic Acids Research 32:
W321-326.
165. Cheng J, Randall AZ, Sweredoski MJ, Baldi P (2005) SCRATCH: a protein structure
and structural feature prediction server. Nucleic Acids Research 33: W72-W76.
166. Lobley A, Sadowski MI, Jones DT (2009) pGenTHREADER and pDomTHREADER:
new methods for improved protein fold recognition and superfamily discrimination.
Bioinformatics 25: 1761-1767.
Bibliography
107
167. Sillitoe I, Cuff AL, Dessailly BH, Dawson NL, Furnham N, et al. (2013) New functional
families (FunFams) in CATH to improve the mapping of conserved functional sites to
3D structures. Nucleic Acids Research 41: D490-498.
168. Jacobson MP, Friesner RA, Xiang Z, Honig B (2002) On the role of the crystal
environment in determining protein side-chain conformations. Journal of Molecular
Biology 320: 597-608.
169. Jacobson MP, Pincus DL, Rapp CS, Day TJ, Honig B, et al. (2004) A hierarchical
approach to all-atom protein loop prediction. Proteins 55: 351-367.
170. Pronk S, Pall S, Schulz R, Larsson P, Bjelkmar P, et al. (2013) GROMACS 4.5: a highthroughput and highly parallel open source molecular simulation toolkit. Bioinformatics
29: 845-854.
171. Van der Spoel D, Lindahl E, Hess B, Groenhof G, Mark AE, et al. (2005) Gromacs:
Fast, Flexible, and Free. Journal of Computational Chemistry 26: 1701-1718.
172. Scott WRP, Hunenberger PH, Tironi IG, Mark AE, Billeter SR, et al. (1999) The
GROMOS biomolecular simulation program package. Journal of Physical Chemistry A
103: 3596-3607.
173. Hess B, Bekker H, Berendsen HJC, Fraaije JGEM (1997) LINCS: A linear constraint
solver for molecular simulations. Journal of Computational Chemistry 18: 1463-1472.
174. Essmann U, Perera L, Berkowitz ML, Darden T, Lee H, et al. (1995) A Smooth Particle
Mesh Ewald Method. Journal of Chemical Physics 103: 8577-8593.
175. Fan H, Mark AE (2004) Refinement of homology-based protein structures by molecular
dynamics simulation techniques. Protein Science : A Publication of the Protein Society
13: 211-220.
176. Elofsson A, Nilsson L (1993) How Consistent Are Molecular-Dynamics Simulations Comparing Structure and Dynamics in Reduced and Oxidized Escherichia-Coli
Thioredoxin. Journal of Molecular Biology 233: 766-780.
177. Mizuguchi K, Blundell T (2000) Analysis of conservation and substitutions of
secondary structure elements within protein superfamilies. Bioinformatics 16: 11111119.
178. Geourjon C, Combet C, Blanchet C, Deléage G (2001) Identification of related proteins
with weak sequence identity using secondary structure information. Protein Science 10:
788-797.
179. Simons KT, Kooperberg C, Huang E, Baker D (1997) Assembly of protein tertiary
structures from fragments with similar local sequences using simulated annealing and
Bayesian scoring functions. Journal of Molecular Biology 268: 209-225.
180. Bradley P, Misura KMS, Baker D (2005) Toward high-resolution de novo structure
prediction for small proteins. Science 309: 1868-1871.
Bibliography
108
181. Kim DE, Chivian D, Baker D (2004) Protein structure prediction and analysis using the
Robetta server. Nucleic Acids Research 32: W526-W531.
182. Michaely P, Tomchick DR, Machius M, Anderson RG (2002) Crystal structure of a 12
ANK repeat stack from human ankyrinR. EMBO J 21: 6387-6396.
183. Malek S, Huxford T, Ghosh G (1998) IκBα Functions through Direct Contacts with
the Nuclear Localization Signals and the DNA Binding Sequences of NF-κB. Journal of
Biological Chemistry 273: 25427-25435.
184. Sue SC, Cervantes C, Komives EA, Dyson HJ (2008) Transfer of flexibility between
ankyrin repeats in I kappa B alpha upon formation of the NF-kappa B complex. Journal
of Molecular Biology 380: 917-931.
185. Huang DB, Huxford T, Chen YQ, Ghosh G (1997) The role of DNA in the mechanism
of NFkappaB dimer formation: crystal structures of the dimerization domains of the p50
and p65 subunits. Structure 5: 1427-1436.
186. Chen FE, Huang DB, Chen YQ, Ghosh G (1998) Crystal structure of p50/p65
heterodimer of transcription factor NF-kappaB bound to DNA. Nature 391: 410-413.
187. Miller S, Janin J, Lesk AM, Chothia C (1987) Interior and Surface of Monomeric
Proteins. Journal of Molecular Biology 196: 641-656.
188. Brockman JA, Scherer DC, McKinsey TA, Hall SM, Qi X, et al. (1995) Coupling of a
signal response domain in I kappa B alpha to multiple pathways for NF-kappa B
activation. Mol Cell Biol 15: 2809-2818.
189. DiDonato JA, Mercurio F, Karin M (1995) Phosphorylation of I kappa B alpha precedes
but is not sufficient for its dissociation from NF-kappa B. Mol Cell Biol 15: 1302-1311.
190. DiDonato J, Mercurio F, Rosette C, Wu-Li J, Suyang H, et al. (1996) Mapping of the
inducible IkappaB phosphorylation sites that signal its ubiquitination and degradation.
Mol Cell Biol 16: 1295-1304.
191. Traenckner EB, Pahl HL, Henkel T, Schmidt KN, Wilk S, et al. (1995) Phosphorylation
of human I kappa B-alpha on serines 32 and 36 controls I kappa B-alpha proteolysis and
NF-kappa B activation in response to diverse stimuli. EMBO J 14: 2876-2883.
192. Whiteside ST, Ernst MK, LeBail O, Laurent-Winter C, Rice N, et al. (1995) N- and Cterminal sequences control degradation of MAD3/I kappa B alpha in response to
inducers of NF-kappa B activity. Mol Cell Biol 15: 5339-5345.
193. Perkins ND (2006) Post-translational modifications regulating the activity and function
of the nuclear factor kappa B pathway. Oncogene 25: 6717-6730.
194. Scheidereit C (2006) I kappa B kinase complexes: gateways to NF-kappa B activation
and transcription. Oncogene 25: 6685-6705.
195. Groban ES, Narayanan A, Jacobson MP (2006) Conformational changes in protein loops
and helices induced by post-translational phosphorylation. Plos Computational Biology
2: 238-250.
Bibliography
109
196. Schrodinger, LLC (2010) The PyMOL Molecular Graphics System, Version 1.3r1.
197. Parrinello M, Rahman A (1981) Polymorphic Transitions in Single-Crystals - a New
Molecular-Dynamics Method. Journal of Applied Physics 52: 7182-7190.
198. Bussi G, Donadio D, Parrinello M (2007) Canonical sampling through velocity
rescaling. Journal of Chemical Physics 126.
199. Baker NA, Sept D, Joseph S, Holst MJ, McCammon JA (2001) Electrostatics of
nanosystems: Application to microtubules and the ribosome. Proceedings of the National
Academy of Sciences 98: 10037-10041.
200. Narayanan A, Jacobson MP (2009) Computational studies of protein regulation by posttranslational phosphorylation. Curr Opin Struct Biol 19: 156-163.
201. Michel Espinoza-Fonseca L, Kast D, Thomas DD (2007) Molecular Dynamics
Simulations Reveal a Disorder-to-Order Transition on Phosphorylation of Smooth
Muscle Myosin. Biophysical Journal 93: 2083-2090.
202. Nurik CE, Cooper DR, Ramaswamy SS, Jayaraman V Effect of Phosphorylation on
Structure of C-Terminal Segment of AMPA Receptor. Biophysical Journal 108: 286a287a.
203. Smart JL, McCammon JA (1999) Phosphorylation stabilizes the N-termini of alphahelices. Biopolymers 49: 225-233.
204. Miranda FF, Teigen K, Thorolfsson M, Svebak RM, Knappskog PM, et al. (2002)
Phosphorylation and mutations of Ser(16) in human phenylalanine hydroxylase. Kinetic
and structural effects. J Biol Chem 277: 40937-40943.
205. Dal Peraro M, Alber F, Carloni P (2001) Ser133 phosphate-KIX interactions in the
CREB-CBP complex: an ab initio molecular dynamics study. Eur Biophys J 30: 75-81.
206. Groban ES, Narayanan A, Jacobson MP (2006) Conformational changes in protein loops
and helices induced by post-translational phosphorylation. PLoS Comput Biol 2: e32.
207. Steinmetz MO, Jahnke W, Towbin H, Garcia-Echeverria C, Voshol H, et al. (2001)
Phosphorylation disrupts the central helix in Op18/stathmin and suppresses binding to
tubulin. EMBO Rep 2: 505-510.
208. Yang J, Cron P, Thompson V, Good VM, Hess D, et al. (2002) Molecular mechanism
for the regulation of protein kinase B/Akt by hydrophobic motif phosphorylation. Mol
Cell 9: 1227-1240.
209. Pons J, Evrard-Todeschi N, Bertho G, Gharbi-Benarous J, Sonois V, et al. (2007)
Structural studies on 24P-IkappaBalpha peptide derived from a human IkappaB-alpha
protein related to the inhibition of the activity of the transcription factor NF-kappaB.
Biochemistry 46: 2958-2972.
210. Evrard-Todeschi N, Pons J, Gharbi-Benarous J, Bertho G, Benarous R, et al. (2008)
Structure of the complex between phosphorylated substrates and the SCF beta-TrCP
Bibliography
110
ubiquitin ligase receptor: a combined NMR, molecular modeling, and docking approach.
J Chem Inf Model 48: 2350-2361.
211. Elcock AH (2004) Molecular simulations of diffusion and association in
multimacromolecular systems. Numerical Computer Methods, Pt D 383: 166-198.
212. Fogolari F, Brigo A, Molinari H (2002) The Poisson-Boltzmann equation for
biomolecular electrostatics: a tool for structural biology. Journal of Molecular
Recognition 15: 377-392.
213. Sheinerman FB, Norel R, Honig B (2000) Electrostatic aspects of protein-protein
interactions. Current Opinion in Structural Biology 10: 153-159.
214. Leaf A, Xiao YF, Kang JX, Billman GE (2003) Prevention of sudden cardiac death by
n-3 polyunsaturated fatty acids. Pharmacology & Therapeutics 98: 355-377.
215. Boland LM, Drzewiecki MM (2008) Polyunsaturated Fatty Acid Modulation of
Voltage-Gated Ion Channels. Cell Biochemistry and Biophysics 52: 59-84.
216. Wilder RM (1921) The effect of ketonemia on course of epilepsy. Mayo Clin Bull 2:
307-308.
217. Borjesson SI, Parkkari T, Hammarstrom S, Elinder F (2010) Electrostatic Tuning of
Cellular Excitability. Biophysical Journal 98: 396-403.
218. Long SB, Campbell EB, MacKinnon R (2005) Voltage sensor of kv1.2: Structural basis
of electromechanical coupling. Science 309: 903-908.
219. Treptow W, Tarek M (2006) Environment of the gating charges in the Kv1.2 Shaker
potassium channel. Biophysical Journal 90: L64-L66.
220. Delemotte L, Tarek M, Klein ML, Amaral C, Treptow W (2011) Intermediate states of
the Kv1.2 voltage sensor from atomistic molecular dynamics simulations. Proceedings of
the National Academy of Sciences of the United States of America 108: 6109-6114.
221. Jensen MO, Jogini V, Borhani DW, Leffler AE, Dror RO, et al. (2012) Mechanism of
voltage gating in potassium channels. Science 336: 229-233.
222. Milescu M, Bosmans F, Lee S, Alabi AA, Il Kim J, et al. (2009) Interactions between
lipids and voltage sensor paddles detected with tarantula toxins. Nature Structural &
Molecular Biology 16: 1080-1085.
223. Mathur R, Zheng J, Yan YY, Sigworth FJ (1997) Role of the S3-S4 linker in Shaker
potassium channel activation. Journal of General Physiology 109: 191-199.
224. Gonzalez C, Rosenman E, Bezanilla F, Alvarez O, Latorre R (2000) Modulation of the
Shaker K+ channel gating kinetics by the S3-S4 linker. Journal of General Physiology
115: 193-207.
225. Eswar N, Webb B, Marti-Renom MA, Madhusudhan MS, Eramian D, et al. (2007)
Comparative protein structure modeling using MODELLER. Curr Protoc Protein Sci
Chapter 2: Unit 2 9.
Bibliography
111
226. Melo F, Sanchez R, Sali A (2002) Statistical potentials for fold assessment. Protein
Science 11: 430-448.
227. Berger O, Edholm O, Jahnig F (1997) Molecular dynamics simulations of a fluid bilayer
of dipalmitoylphosphatidylcholine at full hydration, constant pressure, and constant
temperature. Biophysical Journal 72: 2002-2013.
228. Wolf MG, Hoefling M, Aponte-Santamaria C, Grubmuller H, Groenhof G (2010)
g_membed: Efficient Insertion of a Membrane Protein into an Equilibrated Lipid Bilayer
with Minimal Perturbation. Journal of Computational Chemistry 31: 2169-2174.
229. Lindorff-Larsen K, Piana S, Palmo K, Maragakis P, Klepeis JL, et al. (2010) Improved
side-chain torsion potentials for the Amber ff99SB protein force field. Proteins-Structure
Function and Bioinformatics 78: 1950-1958.
230. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML (1983) Comparison
of Simple Potential Functions for Simulating Liquid Water. Journal of Chemical Physics
79: 926-935.
231. Best RB, Zhu X, Shim J, Lopes PEM, Mittal J, et al. (2012) Optimization of the
Additive CHARMM All-Atom Protein Force Field Targeting Improved Sampling of the
Backbone phi, psi and Side-Chain chi(1) and chi(2) Dihedral Angles. Journal of
Chemical Theory and Computation 8: 3257-3273.
232. Yesselman JD, Price DJ, Knight JL, Brooks CL (2012) MATCH: An Atom-Typing
Toolset for Molecular Mechanics Force Fields. Journal of Computational Chemistry 33:
189-202.
233. Feller SE, Gawrisch K, MacKerell AD (2002) Polyunsaturated fatty acids in lipid
bilayers: Intrinsic and environmental contributions to their unique physical properties.
Journal of the American Chemical Society 124: 318-326.
234. Hess B (2008) P-LINCS: A parallel linear constraint solver for molecular simulation.
Journal of Chemical Theory and Computation 4: 116-122.
235. Henrion U, Renhorn J, Borjesson SI, Nelson EM, Schwaiger CS, et al. (2012) Tracking
a complete voltage-sensor cycle with metal-ion bridges. Proceedings of the National
Academy of Sciences of the United States of America 109: 8552-8557.
236. Knight CJ, Hub JS (Submitted) MemGen: A general web server for the setup of lipid
membrane simulation systems.
237. Xiao YF, Kang JX, Morgan JP, Leaf A (1995) Blocking Effects of Polyunsaturated
Fatty-Acids on Na+ Channels of Neonatal Rat Ventricular Myocytes. Proceedings of the
National Academy of Sciences of the United States of America 92: 11000-11004.
238. Hong MP, Kim HI, Shin YK, Lee CS, Park M, et al. (2004) Effects of free fatty acids on
sodium currents in rat dorsal root ganglion neurons. Brain Research 1008: 81-91.
239. Danthi SJ, Enyeart JA, Enyeart JJ (2005) Modulation of native T-type calcium channels
by omega-3 fatty acids. Biochemical and Biophysical Research Communications 327:
485-493.
Bibliography
112
240. Decher N, Streit AK, Rapedius M, Netter MF, Marzian S, et al. (2010) RNA editing
modulates the binding of drugs and highly unsaturated fatty acids to the open pore of Kv
potassium channels. Embo Journal 29: 2101-2113.
241. Liin SI, Silvera Ejneby M, Barro-Soria R, Skarsfeldt MA, Larsson JE, et al. (2015)
Polyunsaturated fatty acid analogs act antiarrhythmically on the cardiac IKs channel.
Proc Natl Acad Sci U S A 112: 5714-5719.
242. Zheng HF, Li XL, Jin ZY, Sun JB, Li ZL, et al. (2005) Effects of unsaturated fatty acids
on calcium-activated potassium current in gastric myocytes of guinea pigs. World
Journal of Gastroenterology 11: 672-675.
243. Eldho NV, Feller SE, Tristram-Nagle S, Polozov IV, Gawrisch K (2003)
Polyunsaturated docosahexaenoic vs docosapentaenoic acid - Differences in lipid matrix
properties from the loss of one double bond. Journal of the American Chemical Society
125: 6409-6421.
244. Soubias O, Gawrisch K (2007) Docosahexaenoyl chains isomerize on the subnanosecond time scale. Journal of the American Chemical Society 129: 6678-6679.
245. Feller SE (2008) Acyl chain conformations in phospholipid bilayers: a comparative
study of docosahexaenoic acid and saturated fatty acids. Chemistry and Physics of Lipids
153: 76-80.
246. Feller SE, Gawrisch K (2005) Properties of docosahexaenoic-acid-containing lipids and
their influence on the function of rhodopsin. Current Opinion in Structural Biology 15:
416-422.
247. Ottosson NE, Liin SI, Elinder F (2014) Drug-induced ion channel opening tuned by the
voltage sensor charge profile. Journal of General Physiology 143: 173-182.
248. Xu YP, Ramu Y, Shin HG, Yamakaze J, Lu Z (2013) Energetic role of the paddle motif
in voltage gating of Shaker K+ channels. Nature Structural & Molecular Biology 20:
574-+.
249. Kasimova MA, Zaydman MA, Cui JM, Tarek M (2015) PIP2-dependent coupling is
prominent in Kv7.1 due to weakened interactions between S4-S5 and S6. Scientific
Reports 5.
250. Chen L, Zhang Q, Qiu Y, Li Z, Chen Z, et al. (2015) Migration of PIP2 lipids on
voltage-gated potassium channel surface influences channel deactivation. Sci Rep 5:
15079.
251. Pathak MM, Yarov-Yarovoy V, Agarwal G, Roux B, Barth P, et al. (2007) Closing in on
the resting state of the Shaker K(+) channel. Neuron 56: 124-140.
252. Vargas E, Bezanilla F, Roux B (2011) In search of a consensus model of the resting
state of a voltage-sensing domain. Neuron 72: 713-720.
253. Campos FV, Chanda B, Roux B, Bezanilla F (2007) Two atomic constraints
unambiguously position the S4 segment relative to S1 and S2 segments in the closed
Bibliography
113
state of Shaker K channel. Proceedings of the National Academy of Sciences of the
United States of America 104: 7904-7909.
254. Lin MCA, Hsieh JY, Mock AF, Papazian DM (2011) R1 in the Shaker S4 occupies the
gating charge transfer center in the resting state. Journal of General Physiology 138: 155163.
255. Villalba-Galea CA, Sandtner W, Starace DM, Bezanilla F (2008) S4-based voltage
sensors have three major conformations. Proceedings of the National Academy of
Sciences of the United States of America 105: 17600-17607.
256. Schwaiger CS, Bjelkmar P, Hess B, Lindahl E (2011) 3(10)-Helix Conformation
Facilitates the Transition of a Voltage Sensor S4 Segment toward the Down State.
Biophysical Journal 100: 1446-1454.
257. Bezanilla F (2000) The voltage sensor in voltage-dependent ion channels. Physiological
Reviews 80: 555-592.
258. Schwaiger CS, Borjesson SI, Hess B, Wallner B, Elinder F, et al. (2012) The Free
Energy Barrier for Arginine Gating Charge Translation Is Altered by Mutations in the
Voltage Sensor Domain. Plos One 7.
259. Tao X, Lee A, Limapichat W, Dougherty DA, MacKinnon R (2010) A Gating Charge
Transfer Center in Voltage Sensors. Science 328: 67-73.
260. Lacroix JJ, Bezanilla F (2011) Control of a final gating charge transition by a
hydrophobic residue in the S2 segment of a K+ channel voltage sensor. Proceedings of
the National Academy of Sciences of the United States of America 108: 6444-6449.
261. Kosower EM (1985) A Structural and Dynamic Molecular-Model for the SodiumChannel of Electrophorus-Electricus. Febs Letters 182: 234-242.
262. Clayton GM, Altieri S, Heginbotham L, Unger VM, Morais-Cabral JH (2008) Structure
of the transmembrane regions of a bacterial cyclic nucleotide-regulated channel.
Proceedings of the National Academy of Sciences of the United States of America 105:
1511-1515.
263. Shafrir Y, Durell SR, Guy HR (2008) Models of voltage-dependent conformational
changes in NaChBac channels. Biophysical Journal 95: 3663-3676.
264. Bjelkmar P, Niemela PS, Vattulainen I, Lindahl E (2009) Conformational Changes and
Slow Dynamics through Microsecond Polarized Atomistic Molecular Simulation of an
Integral Kv1.2 Ion Channel. Plos Computational Biology 5.
265. Schow EV, Freites JA, Gogna K, White SH, Tobias DJ (2010) Down-State Model of the
Voltage-Sensing Domain of a Potassium Channel. Biophysical Journal 98: 2857-2866.
266. Khalili-Araghi F, Jogini V, Yarov-Yarovoy V, Tajkhorshid E, Roux B, et al. (2010)
Calculation of the Gating Charge for the Kv1.2 Voltage-Activated Potassium Channel.
Biophysical Journal 98: 2189-2198.
Bibliography
114
267. Papazian DM, Shao XM, Seoh SA, Mock AF, Huang Y, et al. (1995) Electrostatic
Interactions of S4 Voltage Sensor in Shaker K+ Channel. Neuron 14: 1293-1301.
115
Acknowledgements
This thesis was made possible with the contributions of many individuals in my life. The people I would like to
give special thanks to are:
Matthias Stein, my supervisor and mentor who has believed in me and encouraged me since day one. Thank
you for giving me the freedom to explore any scientific question I had and offering your full support
throughout this period. I have learned, grown, and flourished into a scientist under your guidance. You are a
true scientist who keeps an open mind and is not afraid of new ideas.
My collaborators in Germany: Prof. Michael Naumann for introducing me into the wonderful world of IκBα
and all the fruitful discussions that followed. Dr. Serdar Durdagi for critical reading of the IκBα paper and
introducing me to protein-protein docking. Prof. Helmut Weiss and Prof. Andreas Seidel-Morgenstern for
welcoming me into the faculty and giving me the chance to do real science in a stimulating environment.
The members of the PCG group at the Max Planck, that made life at the institute and Magdeburg much more
pleasurable, and in their own ways contributed to many happy memories of Germany.
A special thanks to Erik Lindahl who took me under his wings at SciLifeLab and introduced me to the world of
voltage-gated ion channels. I will always remember your encouraging words at the end of every meeting
where you would say ‘Great job!’ no matter what. My collaborators Magnus Andersson for all his help with the PUFA project and for always being optimistic and super efficient. Fredrik Elinder for the invaluable insights he
brought to the PUFA project.
The Lindahl lab members who made life with Gromacs a lot faster and easier. The Carlsson lab members for
all the fun we had exploring Stockholm and introducing me to protein-ligand docking.
The people I met during the course of my PhD with whom I became good friends: Dawid for being the best
office mate in the world and my ally at the Max Planck. Jane and Kasia for all our great talks, trips, and great
laughs we shared together. Narges for all our adventures together, for always making me laugh and
introducing me to the world of TCK. Viveca and Özge for making days on floor 6 a more enjoyable experience,
our movie nights and our scientific discussions.
Felicia for always being there for me no matter where in the world you were. Thank you for being the best
friend I could ever ask for and believing in me. Rachel for being the most awesome friend since day one at
Gillespie’s. Love both of you girls.
Elly and Ila for sharing my DNA and being my partners in crime. This thesis would not have been possible
without you girls. You are my confidantes and make my life richer by simply being the remarkable souls that
you are.
Mom and Dad, I dedicate this thesis to you. You have been the backbone of my life and your faith in me has
been relentless. Thank you for your unconditional love and making me see the importance of education from
an early age.
116
List of Publications
The content of this dissertation contains previously published material in Open Access
journals that permit unrestricted use, distribution, and reproduction in any medium.
I.
II.
Yazdi S, Durdagi S, Naumann M and Stein M (2015) Structural modeling of the Nterminal signal–receiving domain of IκBα. Front. Mol. Biosci.2:32. doi:
10.3389/fmolb.2015.00032.
Yazdi S, Stein M, Elinder F, Andersson M, Lindahl E (2016) The Molecular Basis of
Polyunsaturated Fatty Acid Interactions with the Shaker Voltage-Gated Potassium
Channel. PLoS Comput Biol 12(1): e1004704. doi:10.1371/journal.pcbi.1004704.
Curriculum vitae
Name
Date of birth
Nationality
Address
Samira Yazdi
May 27, 1981
Canadian
Löjtnantsgatan 11, 11550 Stockholm
Education
2011-2016
Ph.D. The Max Planck Institute for Dynamics of Complex Technical Systems,
Germany. Dissertation: The Structural Dynamics of
Soluble and Membrane Proteins Explored through Molecular Simulations
(Supervisor: Matthias Stein)
2007-2009
M.Sc. in Bioinformatics. Stockholm University, Sweden.
Dissertation: Homology modeling and docking help in identifying
pharmacophores in the three human galanin receptor subtypes
(Supervisor: Arne Elofsson)
2001-2005
B.Sc. in Biology and Statistics. University of Toronto, Canada.
2000
Scottish Highers (High School Diploma). James Gillespie’s High School, UK.
Academic experience
Mar 2010 – Jul 2010
Research Assistant at the Molecular Modeling Group,
Department of Biosciences and Nutrition at Karolinska
Institute, Stockholm
Sep 2009 – Feb 2010
Research Assistant at the Unit of Functional Pharmacology,
Department of Neuroscience at Uppsala University, Uppsala
Jun 2009 – Aug 2009
Research Assistant at the Center for Biomembrane Research
(CBR) at the Department of Biophysics and Biochemistry at
Stockholm University, Stockholm